JP2001127999A - Device and method for processing image - Google Patents

Abstract

PROBLEM TO BE SOLVED: To solve a problem that a character written on a color area is not extracted as an image area separation result, since the whole area is extracted by dot concerning the character which is written on the color area such as blue by which the dot is extracted, etc., in image area separation. SOLUTION: A foreground is separated from a background by a binarizing part 11 first concerning input image data and a processing is executed to extract the segment of a width for the portion of prescribed pixels by a segment- extracting part 12. A label part 13 gives the same label to the connected pixels adopted as the foreground in the binarizing part 11, an image feature amount is calculate by a label feature amount calculating part 14 at every label area to which the same label is given and the attribute of the label area is determined by a first attribute determining part 16 from the label feature amount at every label area. Then the segment extracted by the segment extracting part 12 is corrected in a segmentcorrecting part 17 through the use of an attribute determination result by the first attribute determining part 16 at every label area labelled by the labelling part 13.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus such as a copying machine or a facsimile, and a processing method therefor.
The present invention relates to an image processing apparatus and an image processing method that separate and perform image processing suitable for each area.

[0002]

2. Description of the Related Art In an image processing apparatus such as a copying machine or a facsimile, when an input image includes a character area such as a character or a line drawing and a picture area such as a photograph or a picture,
When reproducing the image, it is desirable from the viewpoint of image quality to separate the character area from the picture area, and to perform the processing emphasizing the resolution on the character area and the processing emphasizing the gradation on the picture area. Also in the case of transmitting the image data as described above, it is desirable to separate the character area and the picture area and to perform compression processing on each of them by different methods from the viewpoint of image quality and compression ratio.

[0003] Specifically, an image read by a scanner or the like, or transmitted by facsimile or the like,
When making a hard copy of an image containing a mixture of characters, line drawings, photos, and halftone dots, processing that emphasizes resolution is applied to character regions such as characters and line drawings, and to image regions such as photos and halftone dots. By performing the process with emphasis on gradation, a high-quality reproduced image can be obtained.

When an image including characters, line drawings, photographs, halftone dots, and the like is transferred to an image output device at a remote place via a network, a character area and a picture area are separated from each other. On the other hand, when the compression processing is performed by a different method, high-quality and small-capacity image data can be obtained, so that a high-speed transfer speed and high-quality reproduced image can be obtained.

Here, in order to adaptively process the above-described processing with emphasis on resolution and gradation, depending on each area of an image, it is necessary to accurately separate a character area and a picture area included in the image. is there. Hereinafter, the separation between the character area and the picture area is referred to as image area separation. Regarding this image area separation,
Conventionally, various proposals have been made.

For example, an image is divided into blocks of a certain size, the maximum density and the minimum density of the pixels included in each divided block are determined, and the difference between the maximum density and the minimum density is determined in advance. In comparison with the threshold, a block larger than the threshold is determined as a character area, and a block smaller than the threshold is determined as a picture area (hereinafter referred to as a first related art), or a local operator (filter) centering on a pixel of interest. There is a method of comparing a filtering result with a predetermined threshold, and determining a pixel larger than the threshold as a character area pixel and a pixel smaller than the threshold as a picture area pixel (hereinafter, referred to as a second conventional technique).

Further, as described above, instead of performing image area separation using only one of the first and second prior arts, processing for extracting a character area or processing for extracting a picture area is performed. (Hereinafter, referred to as a third conventional technique) in which a plurality of processes such as the above are performed, and an image area is separated from each processing result with high accuracy.
Has also been proposed (see JP-A-2-294885). This third related art will be briefly described with reference to FIG.

In FIG. 38, the input image signal is
It is supplied to the character extraction unit 101, the halftone dot extraction unit 102, and the contour extraction unit 102. The character extracting unit 101 determines a character area on a pixel-by-pixel basis from an input image signal, and outputs a result of the determination. The halftone dot extraction unit 102 determines a halftone dot area for each pixel, and outputs the determination result. The determination results of the character extracting unit 101 and the halftone dot extracting unit 102 are supplied to the erroneous determination removing unit 105 after the logical OR operation is performed for each pixel in the logical OR operation unit 104. Misjudgment removal unit 1
At 05, the character extraction result is corrected.

Next, the erroneous judgment removing unit 105 and the contour extracting unit 1
Each output result of No. 03 is supplied to the contour regenerating unit 107 after the logical product operation is performed for each pixel in the logical product operation unit 106. In the contour regenerating unit 107, the contour of the character / line image is corrected for the result of the logical product operation of the logical product calculation unit 106. By using the output result of the contour regeneration unit 107 as the image area separation result, highly accurate image area separation can be performed.

[0010]

However, in the case of the first prior art, since the image area separation is performed in block units, the image area determination result of the outline portion of the character / line drawing is faithful to the input image. It is not. Further, in the case of the second prior art, since it is a method of extracting an edge contained in an image, there is a problem that it is easy to extract noise or an edge contained in a picture.

On the other hand, in the case of the above third prior art, there are few erroneous extractions of edges in a picture, and the outline of a character or a line drawing has an extraction result faithful to an input image. Method. However, the characters written on the color background area cannot be extracted. For example, characters written on a color area such as light blue that can be extracted by the halftone dot extraction unit 102 in FIG. There is a problem that a character written on a color region cannot be extracted as an image area separation result because the character is extracted by the character extraction unit 102 or the character extraction unit 101.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems of the prior art. It is an object of the present invention to accurately extract characters and line drawings from an input image regardless of the background color. Another object of the present invention is to provide an image processing apparatus and an image processing method capable of performing high-accuracy image area separation processing without causing erroneous extraction in a picture area.

[0013]

According to the present invention, there is provided an image processing apparatus comprising: input means for inputting image data; and binarization for performing processing for separating a foreground and a background from the image data input by the input means. Means, a line segment extracting means for extracting a line segment having a width of a predetermined number of pixels from the input image data,
Labeling means for assigning the same label to the pixels connected to the foreground by the binarizing means, and for each label area to which the same label has been assigned by the labeling means, the image feature amount of the label area is calculated. Label feature calculation means, attribute determination means for determining the attribute of the label area from the label feature for each label area calculated by the label feature calculation means, and attribute determination result for each label area by the attribute determination means And a line segment correction unit that corrects the line segment extracted by the line segment extraction unit.

In the image processing apparatus having the above configuration, first,
The input unit inputs image data to the binarizing unit and the line segment extracting unit. Then, the binarization means separates the input image data into a foreground and a background, and gives the separation result to a labeling means. The line segment extraction means extracts a line segment having a width of a predetermined number of pixels, and extracts the line segment. The result is given to the line segment correcting means. The labeling means assigns the same label to the connected pixels that have been set to the foreground by the binarization means, and provides the same to the label feature quantity calculating means. Then, for each label area to which the same label has been assigned, the label feature quantity calculating means calculates the image feature quantity of the label area and gives it to the attribute determining means. The attribute determination means determines the attribute of the label area from the label feature amount for each label area, and provides the determination result to the line segment correction means. Then, the line segment correction unit corrects the line segment extracted by the line segment extraction unit using the attribute determination result for each label area.

[0015]

Embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 1 is a block diagram showing an outline of a basic configuration of an image processing apparatus according to the present invention. As apparent from FIG. 1, the image processing apparatus according to the present invention includes an image input unit 1, a data storage unit 2, an arithmetic processing unit 3, a coefficient storage unit 4, a gradation correction unit 5, a color signal conversion unit 6, a space It has a filter unit 7, an image output unit 8, and an image area separation unit 9, and these processing units are connected to each other via a bus line 10.

In the image processing apparatus having the above configuration, the image input section 1 is an image scanner or the like constituted by a photoelectric conversion element such as a CCD image pickup element, and reads an image of a document and converts it into an electric digital image signal. Output. The digital image signal read and converted by the image input unit 1 is an RGB color image signal having a resolution of 400 dpi and 8 bits for each color.
The following is described.

The data storage unit 2 stores image data input by the image input unit 1, image data on which image processing has been performed by each processing unit, processing results calculated by each processing unit, and the like. The arithmetic processing unit 3 performs control of each device and calculation of processing of each unit. The coefficient storage unit 4 stores coefficients necessary for processing of each unit, and the coefficients are read out by an instruction from the arithmetic processing unit 3 or the like.

Data storage unit 2 output from image input unit 1
The image data (8 bits for each color of RGB) stored in is stored in the gradation correction unit 5 in accordance with an instruction from the arithmetic processing unit 3,
The gradation correction section 5 performs gradation correction of the image.
At this time, an input tone correction coefficient is read from the coefficient storage unit 4 according to an instruction from the arithmetic processing unit 3, and tone correction is performed using the input tone correction coefficient. The image data on which the gradation correction has been performed by the gradation correction unit 5 is stored in the data storage unit 2.

The data storage unit 2 output from the gradation correction unit 5
Is read out by the color signal conversion unit 6 according to an instruction from the arithmetic processing unit 3, and the color signal conversion unit 6 converts the RGB image signal into another color signal (for example, L ^* a).
^* b ^* image signal). At that time, the arithmetic processing unit 3
, A color signal conversion coefficient from the RGB image signal to the L ^* a ^* b ^* image signal is read from the coefficient storage unit 4, and the color signal conversion is performed using the color signal conversion coefficient. The image data subjected to the color signal conversion by the color signal conversion unit 6 is stored in the data storage unit 2.

The image data output from the color signal conversion unit 6 and stored in the data storage unit 2, and the image data subjected to image area separation in an image area separation unit 9, which will be described later, are transmitted by an instruction from the arithmetic processing unit 3. L ^* a ^* b while switching the color signal conversion coefficient for each pixel in accordance with the image data read by the color signal conversion unit 6 and subjected to image area separation by the image area separation unit 9.
^* Converted from an image signal to an output color signal (for example, a YMCK image signal). At this time, the color signal conversion coefficient from the L ^* a ^* b ^* image signal to the YMCK image signal is read from the coefficient storage unit 4 according to the instruction of the arithmetic unit 3, and the color signal conversion coefficient is used by using the color signal conversion coefficient. Conversion is performed. The image data subjected to the color conversion by the color signal conversion unit 6 is stored in the data storage unit 2.

The image data output from the color signal conversion unit 6 and stored in the data storage unit 2 and the image data subjected to image area separation in an image area separation unit 9 described later are spatially The spatial filtering process is performed while switching the filtering coefficient for each pixel in accordance with the image data read out by the filter unit 7 and subjected to the image area separation by the image area separation unit 9. At this time, a filtering coefficient is read from the coefficient storage unit 4 according to an instruction from the arithmetic processing unit 3, and a spatial filtering process is performed using the filtering coefficient. The image data subjected to the spatial filtering process by the spatial filter unit 7 is
The data is stored in the data storage unit 2.

The image data output from the spatial filter unit 7 and stored in the data storage unit 2, and the image data subjected to image area separation in the image area separation unit 9, which will be described later, The output gradation correction processing is performed while switching the output gradation correction coefficient for each pixel in accordance with the image data read out by the correction unit 5 and subjected to the image area separation by the image area separation unit 9. At this time, the output gradation correction coefficient is read from the coefficient storage unit 4 according to the instruction of the arithmetic processing unit 3, and the output gradation correction is performed using the output gradation correction coefficient. The image data on which the output gradation correction has been performed by the gradation correction unit 5 is stored in the data storage unit 2.

The data storage unit 2 output from the gradation correction unit 5
Are read out to the image output unit 8 in accordance with an instruction from the arithmetic processing unit 3, and the YMCK image data stored in the
An image is output by, for example, switching a screen or the like for each pixel in accordance with the image data on which the image area separation has been performed by the image area separation unit 9.

The L ^* a ^* b ^* image data output from the color signal conversion unit 6 and stored in the data storage unit 2 is read out to the image area separation unit 9 according to an instruction from the arithmetic processing unit 3, and the image area Separation processing is performed. At this time, an image area separation coefficient is read from the coefficient storage section 4 according to an instruction from the arithmetic processing section 3, and an image area separation process is performed using the image area separation coefficient. The image data subjected to the image area separation by the image area separation unit 9 is stored in the data storage unit 2. Using the image data having undergone image area separation stored in the data storage unit 2, a color signal conversion unit 6, a spatial filter unit 7, a gradation correction unit 5
The processing coefficient in the image output unit 8 is switched for each pixel.

Next, the image area separating section 9 will be described with reference to FIGS.
This will be described with reference to FIG. FIG. 2 is a block diagram illustrating a configuration of the image area separation unit 9 according to the first embodiment of the present invention. FIGS. 3 and 4 are diagrams showing an example of an image change in a process in the image area separating section 9 according to the first embodiment. As is clear from FIG. 2, the image area separating unit 9 according to the first embodiment includes a binarizing unit 11 and a line segment extracting unit 1.
2, labeling unit 13, label feature amount calculating unit 14, small area removing unit 15, first attribute determining unit 16, and line segment correcting unit 1.
7 is provided.

In the image area separating section 9 having the above-described configuration, the binarizing section 11 performs image data stored in the data storage section 2, that is, a resolution of 400 dpi and L ^* a of 8 bits for each pixel.
^* b ^* Image signal is read out and binarized. Or later,
This image data is referred to as L ^* a ^* b ^* original image data. Note that the L ^* a ^* b ^* original image data is
Line segment extraction unit 12 and label feature amount extraction unit 14 to be described later
Even read.

In the binarization unit 11, the read L ^* a ^*
b ^* Extract a foreground area, for example, a character / line segment, a colored area, a photograph / picture area, etc. on a white background from the original image data, and set the extracted pixels to a high level (logic “1”).
A 1-bit pixel signal is output for each pixel in which the unextracted pixels are set to a low level (logic “0”). Taking the case of FIG. 3 as an example, if the image of FIG. 3A is input to the binarization unit 11, a binarized image is output as shown in FIG. 3C. Details of the binarization unit 11 will be described later.

The binarized image signal output from the binarizing section 11 is input to a labeling section 13. Labeling unit 1
In step 3, a process of assigning one label to each region of the extracted pixels to be connected is performed. A well-known technique is used as a labeling method. Referring to FIGS. 3 and 4,
The binarized image shown in FIG. 3C is input to the labeling unit 13, and as shown in FIG.
Labels up to are given. Hereinafter, an image to which a label is added for each connected region is referred to as a labeling image.

The labeling image output from the labeling unit 13 is input to the label feature amount calculation unit 14 and the line segment correction unit 17. The label feature amount calculation unit 14 performs a process of calculating a feature amount for each label region from the labeling image output from the labeling unit 13 and the L ^* a ^* b ^* original image data. Referring to FIGS. 3 and 4, FIG.
The image data with the label shown in FIG. 3A is input to the label feature amount calculation unit 14, and the L ^* a of the image shown in FIG.
^With reference to the ^* b ^* original image data, a feature amount for each label as shown in FIG. 5 is calculated.

In the present embodiment, the number of relevant labels, the number of circumscribed rectangles, the size of circumscribed rectangles, the pixel density (= the number of applicable pixels to be labeled / circumscribed rectangle area), the average value, the variance value, and the class division value are used as the feature values. Although the frequency / class division and the like are calculated, other feature amounts may be calculated and used for the determination process in the first attribute determination unit 16 described later. Details of the label feature amount calculation unit 14 will be described later.

The feature amount for each label calculated by the label feature amount calculating unit 14 is input to the small area removing unit 15 together with the labeling image to which the label is added by the labeling unit 13, and the area size calculated for each label is calculated. A label having an area size equal to or smaller than the predetermined threshold is removed by comparison with a predetermined threshold.

In the case of FIG. 4, if the threshold value is, for example, circumscribed rectangle width = circumscribed rectangle height = 80, the regions of label numbers 4 to N are removed from the labeling image of FIG. A labeling image as shown in (b) is obtained. That is, by determining the area estimated to be a character from the circumscribed rectangle size and removing the area in advance, the processing in the first attribute determination unit 16 described below can be reduced.

The feature quantity of each label from which the small area has been removed, which has been output from the small area removing section 15, is calculated by the first attribute determining section 16.
And the attributes of each label area, such as
A picture area, a uniform color area, and the like are determined. 3 and 4
In the case of, label number 1 and label number 3 are determined to be a uniform color area, and label number 2 is determined to be a picture area. First attribute determination unit 1
Details of 6 will be described later. The attribute determination result for each label area output from the first attribute determination unit 16 is input to the line segment correction unit 17.

On the other hand, the L ^* a ^* b ^* original image data is also input to the line segment extraction unit 12, where characters and line segments contained in the image are extracted. Then, a 1-bit image signal is output for each pixel, in which pixels extracted at this time are at a high level and pixels not extracted are at a low level. In the case of FIG. 3, FIG.
Characters, ruled lines, or line segments existing in the pattern are extracted from the image of FIG. 3 to obtain an image as shown in FIG. Details of the line segment extraction unit 12 will be described later. Line segment extraction unit 12
Are output to the line segment correction unit 17.

In the line segment correction section 17, the labeling image input from the labeling section 13 and the first attribute determination section 1
The line segment extraction image input from the line segment extraction unit 12 is corrected using the attribute determination result for each label area input from the line extraction unit 6. In the case of FIG. 3 and FIG. 4, the photograph / photograph is obtained from the line segment extraction image of FIG. 3B, the labeling image from which the small area is removed in FIG. 4B, and the attribute determination result for each label area.
The line segment existing in the label area determined to be the picture area is removed, and correction is performed so that only the line segment existing in the uniform color area is left. As a result, an image as shown in FIG. Details of the line segment correction unit 17 will be described later. The image data output from the line segment correction unit 17 is output to the image area separation unit 9.
Are stored in the data storage unit 2 as image data subjected to image area separation.

In the above description, the binarizing section 11 sends the labeling section 13, the labeling section 13 sends the label feature calculating section 14 and the line segment correcting section 17, and the label feature calculating section 14 sends the small area removing section. 15, a small area removing unit 15
To the first attribute determination unit 16, the first attribute determination unit 16 to the line segment correction unit 17, and the line segment extraction unit 12 to the line segment correction unit 17.
In the above description, the image data and the calculated feature amount are input. However, the processing result may be stored in the data storage unit 2 and then the processing result may be read to perform the next processing.

The details of the binarizing section 11 will be described with reference to FIG. FIG. 6 is a block diagram illustrating an example of the configuration of the binarization unit 11. The binarization unit 11 according to the present example includes a shadow extraction unit 18, a halftone dot extraction unit 19, and a logical sum operation unit 20.
And the read image data (L
^* a ^* b ^* original image data), the binarization process is performed using only the L ^* original image data.

In the binarizing section 11, the L ^* original image data is inputted to the shadow extracting section 18 and the halftone dot extracting section 19. In the shadow extraction unit 18, from the input image signal,
Areas with high density, such as characters and line segments such as black, blue, and red,
Alternatively, a relatively dense area such as hair, red flowers, green leaves, etc. is extracted even in a picture, and an image signal of 1 bit for each pixel. In this example, the extracted pixel is a high-level signal.
Pixels that are not extracted are output as low-level signals.
The details of the shadow extraction unit 18 will be described later. The image signal output from the shadow extraction unit 18 is input to a logical sum operation unit 20.

The halftone dot extracting section 19 extracts a halftone dot region included in the image from the input image signal.
In this case, a high-level signal is output for an extracted pixel, and a low-level signal is output for a pixel that has not been extracted. The image signal output from the halftone dot extraction unit 19 is input to the logical sum operation unit 20.

As the halftone dot extracting section 19, the one described in the image processing apparatus already proposed by the present inventor (Japanese Patent Laid-Open No.
No. 73503) can be used. Although the details are not described here, the outline is that the input image data is binarized, and the high-level pixels (or the low-level pixels) of the binary image data are obtained.
Is N1 × N1 (for example, N1 =
13) After determining whether or not a periodic structure is formed in the wide area, the halftone dot area is determined using the wide area of N2 × N2 (for example, N2 = 25). It is to extract.

The image data output from the shadow extracting unit 18 and the image data output from the halftone dot extracting unit 19 are both input to a logical sum operation unit 20, and the logical sum operation is performed on the same pixel unit at the same coordinate position. Will be OR operation unit 20
In the above, by performing a logical OR operation on each output image data from the shadow extracting unit 18 and the halftone dot extracting unit 19 on a pixel-by-pixel basis, characters, line segments, etc.
A signal equivalent to the result extracted in step (1) is output from the logical sum operation unit 20.

On the other hand, for the picture area, most of the area is extracted by the halftone dot extracting section 19, but in the area where the density of the hair is extremely high, the periodic structure of the halftone dots is lost and the halftone dot extraction is performed. The part 19 cannot extract it. However, such a region having a very high density can be extracted by the shadow extracting unit 18, and the result of extracting the entire pattern region is output from the logical sum operation unit 20.

Referring to FIG. 7, when an image as shown in FIG. 7A is input to the binarizing unit 11, that is, when it is input to the shadow extracting unit 18 and the halftone extracting unit 19, ,
An image as shown in FIG. 7B is output from the shadow extracting unit 18, and an image as shown in FIG. 7C is output from the halftone dot extracting unit 19. Then, by inputting each image to the logical sum operation unit 20, an image as shown in FIG.

FIG. 8 shows details of the shadow extracting unit 18.
This will be described with reference to FIG. FIG. 8 is a block diagram illustrating an example of the configuration of the shadow extracting unit 18. 8, the image data input to the shadow extracting unit 18 is input to a 3 × 3 pixel average value calculation unit 21 and a 5 × 5 pixel average value calculation unit 22. The 3 × 3 pixel average value calculation unit 21 calculates an average value of 3 × 3 pixels centering on the target pixel. This 3x
The result calculated by the three-pixel average value calculation unit 21 is output to a comparison unit 25 described later, and the comparison unit 25 compares the result with the output result of a limiter 24 described later.

The 5 × 5 pixel average value calculator 22 calculates the average value of 5 × 5 pixels centering on the target pixel. The result calculated by the 5 × 5 pixel average value calculation unit 22 is sent to the addition / subtraction unit 23, and the addition / subtraction unit 23 performs addition / subtraction with a predetermined value VALUE which is arbitrarily set empirically and input separately. . The operation result of the addition / subtraction unit 23 is output to a limiter 24, in which the value is limited between an upper limit value UPPER and a lower limit value LOWER which are arbitrarily set empirically and input separately.

That is, when the operation result of the addition / subtraction unit 23 is larger than the upper limit value UPPER, the upper limit value UPPER
Is smaller than the lower limit LOWER, the lower limit L
Otherwise, the operation result of the addition / subtraction unit 23 is output from the limiter 24 as it is. Limiter 24
Is input to the comparison unit 25, and is compared with the calculation result of the 3 × 3 pixel average value calculation unit 21 described above. When the calculation result of the 3 × 3 pixel average value calculation unit 21 is larger than the output from the limiter 24, a high-level signal is output. When the calculation result is small or equal, a low-level signal is output.

FIG. 9 is a block diagram showing another example of the configuration of the shadow extracting unit 18. In the figure, the same reference numerals are given to the processing parts that perform the same processing as in FIG. In FIG. 9, the image data input to the shadow extraction unit 18 is input to a 3 × 3 pixel average value calculation unit 21, a 5 × 5 pixel average value calculation unit 22, and a selector 26. The 3 × 3 pixel average value calculation unit 21 calculates an average value of 3 × 3 pixels centering on the target pixel. This 3 × 3 pixel average value calculation unit 21
Is output to a comparison unit 25 described later, and the comparison unit 25 compares the calculation result with an output result of a limiter 24 described later. The calculation result of the 3 × 3 pixel average value calculation unit 21 is also input to the selector 26.

The 5 × 5 pixel average value calculator 22 calculates the average value of 5 × 5 pixels centering on the target pixel. The result calculated by the 5 × 5 pixel average value calculation unit 22 is sent to the addition / subtraction unit 23, which performs addition / subtraction with the output result of the addition / subtraction value calculation unit 27 described later.
The operation result of the addition / subtraction unit 23 is output to a limiter 24, where the upper limit value
The value is limited between ER and lower limit LOWER. 5x5
The calculation result of the pixel average value calculation unit 22 is also input to the selector 26.

In the selector 26, the input 3 × 3 pixel average value calculation unit 21, the 5 × 5 pixel average value calculation unit 22 and the output result (target pixel value) of the color signal conversion unit 3 (see FIG. 1)
Is selected by an external input signal (not shown), and the result is input to the addition / subtraction value calculation unit 27. In the addition / subtraction value calculation unit 27, 5
The value to be added or subtracted from the × 5 pixel average value calculation unit 22 is calculated from the output result from the selector 26.

That is, the output value is calculated by adding, subtracting, multiplying, and dividing a plurality of predetermined values with respect to the input value. Also,
The addition / subtraction value calculation unit 27 can be configured by an LUT (Look Up Table). In this case, the output value can be set relatively freely (stored in the LUT) according to the input value. The addition / subtraction unit 23, the limiter 24, and the comparator 25 have been described above, and thus description thereof will be omitted.

As described above, in the shadow extracting section 18, the output result of the 3 × 3 pixel average value calculating section 21 and the limiter 24 are output.
By comparing the output results of (1) and (2), it is possible to extract pixels having a high density and not extract pixels having a low density.
Also, average the pixel values of a relatively narrow area around the pixel of interest,
Since the average value of pixel values in a wider area is compared with a value obtained by performing a predetermined operation, it is possible to extract an area in which the detailed structure of characters and line segments is preserved.

Another example of the binarization unit 11 will be described with reference to the flowchart of FIG. Also in this example, L ^* a ^* b ^{* of the} L ^* a ^* b ^* original image data input to the binarization unit 11 is L ^*.
Binarization processing is performed using the original image data. In this example, the smaller L ^* is, the brighter (whiter), and the larger the L ^* , the darker (black).

Referring to the flowchart of FIG. 10, first, a smoothing process is performed on the entire L ^* original image in order to remove noise and reduce variations at the time of image input (step S101). As the smoothing process, a filtering process using a 3 × 3 pixel filtering coefficient centering on the target pixel may be used. As an example of the filtering coefficient, 1/9 of all 3 × 3 pixels can be considered.

Subsequently, in order to determine a binarization threshold, a histogram of pixel values is created from the smoothed L ^* image data (step S102). FIG. 11 shows an example of a histogram of the generated L ^* image data. When creating a histogram here, a histogram of all pixels may be created, or a histogram may be created while thinning out pixels.

After the histogram of the pixel values is created, subsequently, a binarization threshold is determined from the created histogram. First, in FIG. 11, a pixel value TH10 of a predetermined threshold value
The most frequent pixel value MAX equal to or less than 1 is detected (step S103). Next, the pixel values VAL1 and VAL2 with the least frequency between MAX and TH101 are detected (step S104).

Next, as a result of the detection in step S104, if there are a plurality of pixel values with the least frequency, the smaller one of the pixel values (VAL1) is set as the binarization threshold (step S105). Finally, the L ^* original image data is binarized using the binarization threshold value VAL1 obtained in step S105 (step S106). In this example, a low level signal is output when the L ^* pixel value is smaller than the binarization threshold value VAL1, and a high level signal is output when the L ^* pixel value is equal to or larger than the threshold value VAL1.

The details of the label feature amount calculation unit 14 will be described. The label feature amount calculation unit 14 calculates the feature amount for each label region. In this example, the feature amount is the circumscribed rectangle position, the circumscribed rectangle size, the pixel density (= the number of labeled pixels / the circumscribed rectangle area), and the average value. -Variance value-Class division value frequency-Class variance ratio is calculated. However, since the circumscribed rectangle position, the circumscribed rectangle size, the pixel density, the average value, and the variance value are well known, the description is omitted here, and only the class division value frequency and the class variance ratio will be described.

Also in this example, the label feature amount is calculated using the L ^* a ^* b ^* original image data among the L ^* a ^* b ^* original image data. Note that, as described above, the number of gradations per pixel is 8 bits, that is, the pixel value is represented by 0 to 255. Hereinafter, the class division value frequency / class variance ratio will be described with reference to the flowchart of FIG.

First, a smoothing process is performed on the entire L ^* original image in order to remove noise and reduce variations during image input (step S201). As the smoothing process,
A filtering process using a 3 × 3 pixel filtering coefficient centering on the target pixel may be used. As an example of the filtering coefficient, 1/9 of all 3 × 3 pixels can be considered. Next, a histogram of pixel values is created for each label from the L ^* image data that has been subjected to the smoothing process for calculating the feature amount (step S202). Then, by using the method of the discriminant analysis method, the class division value frequency
Find the class dispersion ratio.

In this discriminant analysis method, first, the maximum var_max of the class variance ratio and the temporary divided pixel value t are initialized (step S203). Then class 0,
That is, the number of pixels p0, the average value ave0 of the pixel values, and the variance var0 of the pixel values between the pixel values 0 to t-1 are calculated (step S204). Next, class 1, that is, when the pixel value is t ~
The pixel number p1, the average value ave1 of the pixel values, and the variance var1 of the pixel values during the period 255 are calculated (step S205).
Next, the number p of pixels, the average ave of pixel values, and the variance var of pixel songs are calculated for the entire target label area (step S2).
06).

Next, the intra-class variance var_W is obtained using the following equation (1) (step S207). var_W = var0 × p0 / p + var1 × p1 / p (1) Then, using the following equation (2), the inter-class variance var_
B is obtained (step S208). var_B = (p0 / p) × (p1 / p) × (ave0−ave1) × (ave0−ave1) (2) Next, using the following equation (3), the class dispersion ratio var_
R is obtained (step S209). var_R = var_B / var_W (3)

Next, the class dispersion ratio var_R is compared with the maximum class dispersion ratio var_max (step S21).
0), the class dispersion ratio var_R is the maximum class dispersion ratio v
If smaller than ar_max, the class dispersion ratio var
The value of _R is the maximum var_max of the class dispersion ratio, and the value of the temporary divided pixel value t is the divided pixel value t that gives the maximum class dispersion ratio.
_Max (step S211).

When the class variance ratio var_R is greater than or equal to the maximum var_max of the class variance ratio, or after passing through step S211, the provisionally divided pixel value t
Is incremented by one (t ← t + 1) (step S212). Then, the temporary divided pixel value t is compared with 256 (step S2).
13) If t <256, the process returns to step S204,
If t ≧ 256, the frequency at the divided pixel value t_max giving the maximum class variance ratio, that is, the class division value frequency f_max is obtained from the histogram (step S21).
4).

In the present embodiment, the feature amount is calculated using only the L ^* image data, but three types of feature amounts are calculated using the L ^* a ^* b ^* image data, or
The feature amount may be calculated dimensionally. In addition, although the feature amount for dividing the histogram into two, class 0 and class 1, is obtained, the feature amount for dividing the histogram into three or more may be obtained.

The details of the first attribute judging section 16 are described in FIG.
This will be described with reference to the flowchart of FIG. The first attribute determination unit 16 determines an attribute for each label area using the feature amount calculated by the label feature amount calculation unit 14. In this example, it is assumed that the number of pixels with corresponding labels, the variance value, the frequency of class division values, and the class variance ratio are used as the feature amounts, and whether the attribute is a photograph / pattern region or a uniform color region is determined as the attribute.

Referring to the flowchart of FIG. 13, first, the variance value of the label area is set to a predetermined threshold value TH.
201 (step S301). At this time, when it is determined that the variance value is greater than or equal to the threshold value TH201, the class variance ratio of the corresponding label area is set to:
A comparison is made with a predetermined threshold value TH202. If it is determined that the label area is small, the attribute of the label area is determined to be a uniform color area (step S303).

If it is determined in step S302 that the class dispersion ratio of the corresponding label area is greater than or equal to the threshold value TH202, the class division value frequency of the corresponding label area is set to a predetermined coefficient T
The value of the label area is compared with the value obtained by multiplying by H203 (step S304).
It is determined as a picture area (step S305).

If it is determined in step S304 that the class division value frequency of the corresponding label area is smaller than the value obtained by multiplying the number of pixels having the corresponding label by the coefficient TH203, the flow advances to step S303 to change the attribute of the relevant label area. If it is determined that the label region is a uniform color region and it is determined that the label region is large or equal, the process advances to step S305 to determine that the attribute of the label region is a photograph / picture region.

After the processing in step S303 or S305, it is determined whether or not the attribute determination for all the label areas has been completed (step S306). Is completed, and if there is a label area for which attribute determination has not been performed, the process returns to step S301, and the above-described series of determination processing is repeated.

As described above, the first attribute judging section 16 judges the attribute of the corresponding label area using the variance, the class division value frequency, and the class variance for each label. That is, when the variance value is small, it is considered that the corresponding label area is composed of substantially a single color, so that the label area is determined to be a uniform color area. Even if the variance value is large, if the class variance ratio is large and the frequency at which the histogram is divided is small, the histogram has a bimodal shape. -Since the area is considered to be a line-drawn area, the area is determined to be a uniform color area. Otherwise, it is determined to be a photo / picture area.

Next, details of the line segment extraction unit 12 will be described. In FIG. 14 showing an example of the configuration, the image data input to the line segment extraction unit 12 is divided into a first reduction unit 30-
The image data is input to the first to N-th reduction units 30-N, and the image data is reduced. For example, in the first reduction unit 30-1, the image is reduced by half in both the vertical and horizontal directions. In the second reduction unit 30-2, the image is reduced by 1/4 in both the vertical and horizontal directions. Similarly, the N-th reduction unit 3
In 0-N, the image is reduced by 1 / (2N) in both the vertical and horizontal directions. 1st reduction unit 30-1 to Nth reduction unit 30
As the image reduction method in −N, a simple thinning method, 4
A well-known reduction method such as an inter-point interpolation method, a 16-point interpolation method, a projection method, and a median value adoption method may be used.

The first reducing section 30-1 to the N-th reducing section 30-N
The image data reduced by the line extraction unit 31-1
To the line extraction unit 31-N. In addition, the line segment extraction unit 1
2 is also input to the line extraction unit 31-0. In the line extraction units 31-0 to 31-N,
A line segment having a predetermined range of width is extracted from the input image data. In the case of this example, the first reduction units 30-1 to 30-1
Each of the image data is divided by 1 / N by the N-th reduction unit 30-N.
Since the reduction is performed to 2, 1/4, 1/6,..., 1 / (2N), the line extraction units 31-0 to 31-N generate 1
A pixel and a line segment having a width of 2 pixels are extracted. Line extraction unit 31-
Details of the 0 to line extraction unit 31-N will be described later.

On the other hand, the image data input to the line segment extraction unit 12 is also input to the edge extraction unit 32. The edge extraction unit 32 extracts edge pixels included in the image data, and outputs the edge extraction results to the first enlargement unit 33-1 to the N-th enlargement unit 33-N. The details of the edge sleeve portion 32 will be described later.

The image data output from the line extracting units 31-1 to 31-N are respectively stored in the first expanding units 33-1 to N-th.
The image data output from the enlargement unit 33-N and output from the edge extraction unit 32 are also input to the first enlargement unit 33-1 to the N-th enlargement unit 33-N. In the first enlargement unit 33-1 to the N-th enlargement unit 33-N, based on the image data input from the edge extraction unit 32, the line extraction units 31-1 to 31-N
The image data input from N is enlarged. The enlargement ratio in the first enlargement unit 33-1 to the N-th enlargement unit 33-N is the reciprocal of the reduction ratio in the first reduction unit 30-1 to the N-th reduction unit 30-N, that is, 2,4,6
.., 2N values. The details of the first to third enlargement units 33-1 to 33-N will be described later.

The first expanding section 33-1 to the N-th expanding section 33-N
The image data of the output result from the above-described line extraction unit 31-0 is input to the logical sum operation unit 34, and the logical sum operation is performed for each pixel. The image data of the output result from the OR operation unit 34 is the output result of the line segment extraction unit 12.

Subsequently, the line extracting units 31-0 to 31-3
1-N will be described. As mentioned above, in this example,
Although the processing is to extract a line segment having a width of one pixel and two pixels, the processing is not limited to this size, and the processing of extracting other line widths is not limited. In the description here, the line extraction unit 31
It is assumed that all of the line extraction units 31-0 to 31-N are the same processing, and only the line extraction unit 31-0 will be described. You may do it. FIG. 15 is a block diagram illustrating an example of a configuration of the line extracting unit 31-0.

In FIG. 15, the image signal input to the line extracting unit 31-0 is a one-pixel horizontal line detecting unit 41 for detecting a horizontal line having a line width of one pixel, and detecting a vertical line having a line width of one pixel. One-pixel vertical line detector 42, one-pixel left diagonal line detector 43 that detects an upward diagonal line having a line width of one pixel, and one-pixel right diagonal line detection that detects a diagonally right upward line having a line width of one pixel Unit 44, a two-pixel horizontal line detection unit 45 that detects a horizontal line having a line width of two pixels, a two-pixel vertical line detection unit 46 that detects a vertical line having a line width of two pixels, and an ascending left line having a line width of two pixels Is input to a two-pixel right diagonal line detection unit 48 that detects a diagonally rising line with a line width of two pixels.

One pixel horizontal line detector 41, one pixel vertical line detector, one pixel left oblique line detector 43, one pixel right oblique line detector 4
4, 2 pixel horizontal line detection unit 45, 2 pixel vertical line detection unit 46, 2
The details of the pixel left diagonal line detection unit 47 and the two pixel right diagonal line detection unit 48 will be described later. Detected results of horizontal, vertical, left-upward diagonal lines and right-upward diagonal lines and 1-pixel-wide horizontal, vertical, upward-leftward diagonal lines, and right-upward diagonal lines detected by the detection units 41 to 48, respectively. Is the logical sum operation unit 4
9 and a logical sum operation of the detection result is performed for each pixel. Then, the result of the logical sum operation performed by the logical sum operation unit 49 is output as the line extraction result by the line extraction unit 31-0.

Next, regarding the one-pixel horizontal line detection unit 41,
This will be described with reference to FIGS. FIG. 16 is a diagram illustrating the types of lines detected by each line detection unit having a width of one pixel. The one-pixel horizontal line detection unit 41 detects a one-pixel-wide horizontal line as shown in FIG. 16B using a 3 × 3 pixel matrix. FIG. 17 illustrates an example of the configuration of the one-pixel horizontal line detection unit 41.
, The input image data includes a first row average value calculation section 51, a second row average value calculation section 52, a third row average value calculation section 53, a second row minimum value calculation section 54, and a first row maximum value. Calculation unit 5
5, and is input to the third row maximum value calculation unit 56, respectively.

The first row average value calculation unit 51 calculates the average of the pixel values of the three pixels in the first row of the 3 × 3 pixel matrix, that is, FIG.
The average of the pixel values of A, B, and C shown in FIG. , The pixel value of each pixel as _{P A, P B, P C} , average _{AVE (P A, P B,} P C) expressed as. In the second row average value calculation unit 52, AVE (P _D , P _E , P _F )
AVE (P _G , P _H ,
P _I ).

In the second row minimum value calculating section 54, D, E, F
Is obtained. From now on, the minimum value of a plurality of pixels will be expressed as MIN (P _D , P _E , P _F ). The first row maximum value calculation unit 55 determines the maximum value of the pixel values of A, B, and C. In the future, the maximum value of a plurality of pixels will be MAX (P _A , P _B , P
_C ). Further, the third row maximum value calculation section 56 calculates MAX (P _G , P _H , P _I ).

The calculation result of the second row average value calculation section 52 and the first
The calculation results of the row average value calculation unit 51 are both input to the subtraction unit 57-1, and the first row
A process of subtracting the calculation result of the row average value calculation unit 51 is performed. The calculation result of the second row average value calculation section 52 and the calculation result of the third row average value calculation section 53 are both input to the subtraction section 57-2, and the third row average value calculation section 52 calculates the third A process of subtracting the calculation result of the row average value calculation unit 53 is performed.

The operation result of the subtraction section 57-1 is compared with the comparison section 58.
-1 is input and compared with a predetermined threshold value TH1.
The high level is output from the comparison section 58-1 when the operation result of the subtraction section 57-1 is larger than the threshold value TH1, and the low level is output when the result is equal to or smaller than the threshold value TH1. Also,
The operation result in the subtraction unit 57-2 is input to the comparison unit 58-2, and is compared with a predetermined threshold value TH1, and the subtraction unit 57-2
The high level is output from the comparing unit 58-2 when the result of the calculation at 2 is larger than the threshold value TH1, and the low level is output when the result is equal to or smaller than the threshold value TH1.

The calculation result of the second row minimum value calculation section 54 is input to the comparison section 58-3, and is compared with a predetermined threshold value TH2. Is larger than the threshold value TH2, the high level is output from the comparison unit 58-3 when the value is equal to or smaller than the threshold value TH2.

The calculation result of the first row maximum value calculation unit 55 is input to the comparison unit 58-4, and is compared with a predetermined threshold value TH3. Is smaller than the threshold value TH3, the high level is output from the comparing section 58-4 if the level is larger or equal. The calculation result in the third row maximum value calculation unit 56 is input to the comparison unit 58-5 and compared with a predetermined threshold value TH3, and the calculation result in the third row maximum value extraction unit 56 is The high level is output from the comparing section 58-5 when the level is smaller than the threshold value TH3, and the low level is output when the level is higher or equal to the threshold value TH3.

All the comparison results of the comparison units 58-1 to 58-5 are input to the AND operation unit 59, and the AND operation unit 59
Then, the logical product of the comparison results of the comparison units 58-1 to 58-5 is calculated. The result output from the AND operation unit 59 is the output result of the one-pixel horizontal line detection unit 41. When the result is high, it is determined that there is a one-pixel width horizontal line as shown in FIG. In the case of the level, it is determined that there is no horizontal line of one pixel width.

That is, in the one-pixel horizontal line detecting section 41, when all of the following equations (1) to (5) are satisfied, FIG.
It is determined that there is a horizontal line having a width of one pixel as shown in FIG. 16B, and high levels are output as pixel values corresponding to D, E, and F in FIG. If at least one of (5) is not satisfied, it is determined that there is no horizontal line having a width of one pixel as shown in FIG.

AVE (P _D , P _E , P _F ) −AVE (P _A , P _B , P _C )> TH1 (1) AVE (P _D , P _E , P _F ) −AVE (P _G , P _H , P _I )> TH 1 (2) MIN (P _D , P _E , P _F )> TH 2 (3) MAX (P _A , P _B , P _C ) <TH 3 (4) MAX (P _G , P _H , P _I ) <TH3 (5)

Next, the one-pixel vertical line detector 42 will be described. The one-pixel vertical line detection unit 42 detects a one-pixel width vertical line as shown in FIG. 16C using a 3 × 3 pixel matrix. FIG. 18 shows a configuration example of the one-pixel vertical line detection unit 42. The configuration is the same as the one-pixel horizontal line detection unit 41 shown in FIG.
In contrast, the first row average value calculation section 51, the second row average value calculation section 52, the third row average value calculation section 53, the second row minimum value calculation section 54, the first row maximum value calculation section 55, and the The three-row maximum value calculation unit 56 includes a first column average value calculation unit 61, a second column average value calculation unit 62, a third column average value calculation unit 63, a second column minimum value calculation unit 64, and a first column maximum value. Only the calculation unit 65 and the third column maximum value calculation unit 66 are replaced with each other, and are basically the same.

Therefore, here, the description will be made using equations. In the one-pixel vertical line detection unit 42, the following equations (6) to
When all the expressions (10) are satisfied, it is determined that there is a vertical line having a width of one pixel as shown in FIG.
When a high level is output as a pixel value corresponding to B, E, and H, and when even one of (Equation 6) to (Equation 10) does not hold, one pixel as shown in FIG. It is determined that there is no vertical vertical line, and a low level is output for each.

[0092] _{AVE (P B, P E,} P H) one _{AVE (P A, P D,} P G)> TH4 ... (6) AVE (P B, P E, P H) -AVE (P C, P _F , P _I )> TH 4 (7) MIN (P _B , P _E , P _H )> TH 5 (8) MAX (P _A , P _D , P _G ) <TH 6 (9) MAX (P _C , P _F , P _I ) <TH6 (10)

Next, the one-pixel left-slanting-line detector 43 will be described. The one-pixel left diagonal line detection unit 43 detects an upward diagonal line with a one-pixel width as shown in FIG. 16D using a 3 × 3 pixel matrix. FIG. 19 shows one-pixel vertical line detection unit 4.
2 shows a configuration example. The configuration is different from the one-pixel horizontal line detection unit 41 of FIG. 17 in that a first row average value calculation unit 51, a second row average value calculation unit 52, a third row average value calculation unit 53, and a second row minimum value calculation unit The unit 54, the first row maximum value calculation unit 55, and the third row maximum value calculation unit 56 include a lower left three pixel average value calculator 71, an upper left and lower right diagonal three pixel average value calculator 72, and an upper right three pixel average value Only the calculation unit 73, the upper left / lower right diagonal three-pixel minimum value calculator 74, the lower left three-pixel maximum value calculator 75, and the upper right three-pixel maximum value calculator 76 are replaced. A detailed description is omitted.

Therefore, here, the description will be made using equations. In the one-pixel left-slanting-line detecting unit 43, the following equation (1)
When all of the expressions (1) to (15) hold, FIG.
It is determined that there is a one-pixel-wide diagonal line as shown in FIG. 16D, and high levels are output as pixel values corresponding to A, E, and I in FIG. ~ Equation (1
If at least one of 5) is not satisfied, it is determined that there is no oblique line rising to the left of one pixel width as shown in FIG.

AVE (P _A , P _E , P _I ) −AVE (P _D , P _G , P _H )> TH7 (11) AVE (P _A , P _E , P _I ) −AVE (P _B , P _{C, P F)> TH7 ...} (12) MIN (P A, P E, P I)> TH8 ... (13) MAX (P D, P G, P H) <TH9 ... (14) MAX (P B, P _C , P _F ) <TH9 (15)

Next, the one-pixel oblique line detector 44 will be described. The one-pixel right diagonal line detection unit 44 detects an upward diagonal line with a one-pixel width as shown in FIG. 16E using a 3 × 3 pixel matrix. FIG. 20 shows a configuration example of the one-pixel oblique line detection unit 44. The configuration is different from the one-pixel horizontal line detection unit 41 of FIG. 17 in that a first row average value calculation unit 51, a second row average value calculation unit 52, a third row average value calculation unit 53, and a second row minimum value calculation unit The unit 54, the first row maximum value calculation unit 55, and the third row maximum value calculation unit 56 include an upper left three pixel average value calculator 81, an upper right and lower left diagonal three pixel average value calculator 82, and a lower right three pixel average value. Only the calculation unit 83, the upper right / lower left diagonal three-pixel minimum value calculator 84, the upper left three-pixel maximum value calculator 85, and the lower right three-pixel maximum value calculator 86 are replaced. A detailed description is omitted.

Therefore, here, the description will be made using equations. The one-pixel oblique line detection unit 44 calculates the following equation (1)
When all the expressions (6) to (20) hold, FIG.
It is determined that there is an oblique line rising to the right with a width of one pixel as shown in (e), and a high level is output as a pixel value corresponding to C, E, G in FIG. ~ Equation (2
If at least one of 0) does not hold, it is determined that there is no oblique line rising to the right with one pixel width as shown in FIG.

AVE (P _C , P _E , P _G ) −AVE (P _A , P _B , P _D )> TH10 (16) AVE (P _C , P _E , P _G ) −AVE (P _F , P _G ) _H , P _I )> TH 10 (17) MIN (P _C , P _E , P _G )> TH 11 (18) MAX (P _A , P _B , P _D ) <TH 12 (19) MAX (P _F , P _H , P _I ) <TH13 (20)

Next, regarding the two-pixel horizontal line detection unit 45,
This will be described with reference to FIGS. FIG. 21 is a diagram illustrating the types of lines detected by each line detection unit having a width of two pixels. The two-pixel horizontal line detection unit 45 detects a horizontal line having a width of two pixels as shown in FIG. 21B using a 4 × 4 pixel matrix. FIG. 22 showing an example of the configuration of the two-pixel horizontal line detection unit 45.
, The input image data is divided into a first row average value calculation section 91, a second and third row average value calculation section 92, a fourth row average value calculation section 93, a second and third row minimum value calculation section 94, It is input to the one-row maximum value calculation unit 95 and the fourth-row maximum value calculation unit 96, respectively.

In the first row average value calculation section 91, the average of the pixel values of the four pixels in the first row of the 4 × 4 pixel matrix, that is, FIG.
The average AVE (P) of the pixel values of J, K, L and M shown in FIG.
_J, P _K, P _L, the P _M) seek. In the second / third row average value calculation unit 92, AVE (P _N , P _O , P _P , P _Q ,
P _R , P _S , P _T , P _U ) by the fourth row average value calculation unit 93
Then, AVE (P _V , P _W , P _X , P _Y ) is obtained.

In the second / third row minimum value calculating section 94, N,
Minimum value MIN of each pixel value of O, P, Q, R, S, T, U
(P _N , P _O , P _P , P _Q , P _R , P _S , P _T , P _U )
Ask for. In the first row maximum value calculation unit 95, J, K, L,
The maximum value MAX (P _J , P _K , P _L ,
P _M ). In the fourth row maximum value calculation unit 96,
MAX (P _V , P _W , P _X , P _Y ) is obtained.

The calculation result of the second and third row average value calculation section 92 and the calculation result of the first row average value calculation section 91 are both subtracted by the subtraction section 97-
1 and a process of subtracting the calculation result of the first row average value calculation unit 91 from the calculation result of the second and third row average value calculation unit 92 is performed. Both the calculation result of the second and third row average value calculation section 92 and the calculation result of the fourth row average value calculation section 93 are input to the subtraction section 97-2, and the calculation of the second and third row average value calculation section 92 is performed. A process of subtracting the calculation result of the fourth row average value calculation unit 93 from the result is performed.

The operation result of subtraction section 97-1 is compared with comparison section 98.
-1 is input and compared with a predetermined threshold value TH13. When the operation result of the subtraction unit 97-1 is larger than the threshold value TH13, the high level is set. Is output from the comparing section 98-1.
Also, the operation result of the subtraction unit 97-2 is input to the comparison unit 98-2 and compared with a predetermined threshold value TH13, and the operation result of the subtraction unit 97-2 is larger than the threshold value TH13. The high level is output from the comparator 98-2 when the level is equal or lower.

The calculation result of the second / third row minimum value calculating section 94 is input to the comparing section 98-3, and is compared with a predetermined threshold value TH14. When the operation result is larger than the threshold value TH14, the high level is output from the comparison unit 98-3 when the operation result is equal to or smaller than the threshold value TH14.

The calculation result of the first row maximum value calculation unit 95 is input to the comparison unit 98-4, where it is compared with a predetermined threshold value TH15. Is smaller than the threshold value TH15, the high level is output from the comparing section 98-4 when the level is larger or equal. The calculation result in the fourth row maximum value calculation unit 96 is input to the comparison unit 98-5, and is input to the predetermined threshold TH15.
The high level is obtained when the operation result of the fourth row maximum value calculation unit 96 is smaller than the threshold value TH15, and the low level is obtained when the operation result is larger or equal to the threshold value. Is output.

All the comparison results of the comparison units 98-1 to 98-5 are input to the AND operation unit 99, and the AND operation unit 99
, The logical product of the comparison results of the comparison units 98-1 to 98-5 is calculated. The result output from the logical product operation unit 99 is the output result of the two-pixel horizontal line detection unit 45. When the result is high, it is determined that there is a two-pixel width horizontal line as shown in FIG. In the case of the level, it is determined that there is no horizontal line having a width of two pixels.

That is, when all of the following equations (21) to (25) hold, the two-pixel horizontal line detector 45 determines that there is a horizontal line having a two-pixel width as shown in FIG. A high level is output as a pixel value corresponding to N, O, P, Q, R, S, T and U in FIG. 21 (a), and conversely, even one of equations (21) to (25) holds. If not,
It is determined that there is no horizontal line having a width of two pixels as shown in FIG.

AVE (P _N , P _O , P _P , P _Q , P _R , P _S , P _T , P _U ) −AVE (P _J , P _K , P _L , P _M )> TH13 (21) AVE (P _N , P _O , P _P , P _Q , P _R , P _S , P _T , P _U ) −AVE (P _V , P _W , P _X , P _Y )> TH 13 (22) MIN (P _{N, P O, P P,} P Q, P R, P S, P T, P U)> TH14 ... (23) MAX (P J, P K, P L, P M) <TH15 ... (24) MAX (P _V , P _W , P _X , P _Y ) <TH15 (25)

Next, the two-pixel vertical line detector 46 will be described. The two-pixel vertical line detector 46 detects a two-pixel vertical line as shown in FIG. 21C using a 4 × 4 pixel matrix.

FIG. 23 shows an example of the configuration of the two-pixel vertical line detector 46. The configuration is different from the two-pixel horizontal line detection unit 45 in FIG. 22 in that a first row average value calculation unit 91, a second and third row average value calculation unit 92, a fourth row average value calculation unit 93, a second and third The row minimum value calculation unit 94, the first row maximum value calculation unit 95, and the fourth row maximum value calculation unit 96 include a first column average value calculation unit 101,
Column average value calculation section 102, fourth column average value calculation section 103, second and third column minimum value calculation section 104, first column maximum value calculation section 10
Only the fifth and fourth column maximum value calculation units 106 have been replaced, and are basically the same.

Therefore, here, the description will be made using equations. In the two-pixel vertical line detector 46, the following equation (26) is used.
When all of the expressions (30) are satisfied, it is determined that there is a vertical line having a width of two pixels as shown in FIG.
A high level is output as a pixel value corresponding to K, L, O, P, S, T, W, and X in FIG.
When even one of the expressions (30) is not satisfied, FIG.
It is determined that there is no vertical line having a width of two pixels as shown in (c), and a low level is output for each.

AVE (P _K , P _L , P _O , P _P , P _S , P _T , P _W , P _X ) −AVE (P _J , P _N , P _R , P _V )> TH 16 (26) AVE (P _K , P _L , P _O , P _P , P _S , P _T , P _W , P _X ) −AVE (P _M , P _Q , P _U , P _Y )> TH 16 (27) MIN (P _K , P _L , P _O , P _P , P _S , P _T , P _W , P _X )> TH 17 (28) MAX (P _J , P _N , P _R , P _V ) <TH 18 (29) MAX (P _M , P _Q , P _U , P _Y ) <TH18 (30)

Next, the two-pixel left oblique line detecting section 47 will be described. The two-pixel left oblique line detection unit 47 detects an obliquely upward diagonal line having a two-pixel width as shown in FIG. 21D using a 4 × 4 pixel matrix.

FIG. 24 shows an example of the configuration of the two-pixel left oblique line detecting section 47. The configuration is the same as that of the two-pixel horizontal line detector 45 shown in FIG.
The first row average value calculation section 91, the second and third row average value calculation section 92, the fourth row average value calculation section 93, the second and third row minimum value calculation section 94, the first row maximum value calculation The unit 95 and the fourth row maximum value calculation unit 96 include a lower left three pixel average value calculator 111, an upper left / lower right lower 10 pixel average value calculator 112, an upper right three pixel average value calculator 113, an upper left / lower right tilt Only the direction 10 pixel minimum value calculation unit 114, the lower left 3 pixel maximum value calculation unit 115, and the upper right 3 pixel maximum value calculation unit 116 are replaced.
The details are omitted because they are basically the same.

Therefore, here, description will be made using an equation. In the two-pixel left oblique line detection unit 47, the following equation (3)
When all of 1) to (35) hold, FIG.
It is determined that there is a two-pixel-width oblique line rising to the left as shown in FIG. 21D, and JKNOPSTST in FIG.
A high level is output as a pixel value corresponding to U, XY, and Y. On the contrary, when even one of Expressions (31) to (35) does not hold, a two-pixel width as shown in FIG. It is determined that there is no oblique line rising to the left, and each outputs a low level.

AVE (P _J , P _K , P _N , P _O , P _P , P _S , P _T , P _U , P _X , P _Y ) −AVE (P _R , P _V , P _W )> TH 19 ... _{(31) AVE (P J,} P K, P N, P O, P P, P S, P T, P U, P X, P Y) -AVE (P L, P M, P Q)> TH19 ... _{(32) MIN (P J,} P K, P N, P O, P P, P S, P T, P U, P X, P Y)> TH20 ... (33) MAX (P R, P V, P _{W) <TH21 ... (34)} MAX (P L, P M, P Q) <TH21 ... (35)

Next, the two-pixel oblique line detection unit 48 will be described. The two-pixel right diagonal line detection unit 48 detects a two-pixel width diagonally rightward diagonal line as shown in FIG. 21E using a 4 × 4 pixel matrix.

FIG. 25 shows a configuration example of the two-pixel oblique line detection unit 48. The configuration is the same as that of the two-pixel horizontal line detector 45 shown in FIG.
The first row average value calculation section 91, the second and third row average value calculation section 92, the fourth row average value calculation section 93, the second and third row minimum value calculation section 94, the first row maximum value calculation The unit 95 and the fourth row maximum value calculation unit 96 include an upper-left three-pixel average value calculator 121, an upper-right / lower-left lower direction 10-pixel average value calculator 122, a lower-right three-pixel average value calculator 123, and an upper-right / lower-left lower direction Only the 10 pixel minimum value calculation unit 124, the upper left 3 pixel maximum value calculation unit 125, and the lower right 3 pixel maximum value calculation unit 126 are replaced.
The details are omitted because they are basically the same.

Therefore, here, the description will be made using equations. The two-pixel oblique line detection unit 48 calculates the following equation (3)
When all of the expressions 6) to (40) hold, FIG.
It is determined that there is a two-pixel-width upward diagonal line as shown in (e), and L, M, O, P, Q, R, S, and L in FIG.
A high level is output as a pixel value corresponding to TVW. Conversely, if at least one of Equations (36) to (40) does not hold, a two-pixel width as shown in FIG. It is determined that there is no oblique line rising to the right, and each outputs a low level.

AVE (P _L , P _M , P _O , P _P , P _Q , P _R , P _S , P _T , P _V , P _W )-AVE (P _J , P _K , P _N )> TH 22 ... (36) AVE (P _L , P _M , P _O , P _P , P _Q , P _R , P _S , P _T , P _V , P _W )-AVE (P _U , P _X , P _Y )> TH22 ... (37) MIN (P _L , P _M , P _O , P _P , P _Q , P _R , P _S , P _T , P _V , P _W )> TH23 (38) MAX (P _J , P _K , P _{N) <TH24 ... (39)} MAX (P U, P X, P Y) <TH24 ... (40)

Subsequently, the line extracting units 31-0 to 3-3
Another example of 1-N will be described. As described above, in the present embodiment, the line segment having a width of one pixel and two pixels is extracted. However, the present invention is not limited to this size, and may be a process of extracting another line width. In this description, it is assumed that all the line extraction units 31-0 to 31-N perform the same processing, and only the line extraction unit 31-0 will be described. Lines may be extracted as different processes in the units 31-N. FIG.
FIG. 14 is a block diagram illustrating another example of the configuration of the line extraction unit 31-0.

In FIG. 26, the image signal input to the line extracting unit 31-0 includes a one-pixel horizontal line left detecting unit 131 and a one-pixel horizontal line right detecting unit 13 for detecting a horizontal line having a line width of one pixel.
2, a one-pixel vertical line upper detector 133 for detecting a vertical line having a line width of one pixel and a one-pixel vertical line lower detector 134 for detecting a left-upward diagonal line having a line width of one pixel; Detecting unit 135 and one-pixel left-below-line detecting unit 136, one-pixel right-behind-line detecting unit 1 for detecting a right-upward diagonal line having a line width of one pixel
37 and 1 pixel right diagonal under line detection section 138 and 2 pixel horizontal line left detection section 139 and 2 pixel horizontal line right detection section 140 which detect a horizontal line having a line width of 2 pixels. A vertical line having a line width of 2 pixels is detected. Two-pixel vertical line above detector 141 and two-pixel vertical line below detector 14
A two-pixel left diagonal upper detecting unit 143 and a two-pixel left diagonal lower detecting unit 14 for detecting a diagonally rising line having a line width of 2, 2 pixels
4, a two-pixel right oblique line upper detecting unit 145 and a two-pixel right oblique line lower detecting unit 146 for detecting an upward diagonal line having a line width of two or two pixels
Respectively.

One-pixel horizontal line left detector 131, one-pixel horizontal line right detector 132, one-pixel vertical line upper detector 133, one-pixel vertical line lower detector 134, one-pixel upper left line detector 135, one-pixel lower left line A detection unit 136, a detection unit 137 on the right diagonal line of one pixel,
One pixel right diagonal line under detection unit 138, two pixel horizontal line left detection unit 13
9, 2 pixel horizontal line right detector 140, 2 pixel vertical line detector 1
41, a two-pixel vertical line under detection unit 142, a two-pixel left diagonal under-detection unit 143, a two-pixel left diagonal under-detection unit 144, a two-pixel right diagonal under-detection unit 145, and a two-pixel right diagonal under-line detection unit 146 It will be described later.

The horizontal / vertical / left-upward diagonal lines, rightward-upward diagonal lines, 1-pixel-wide horizontal / vertical / leftward-upward diagonal lines, and rightward-upward diagonal lines detected by the detection units 131 to 146, respectively. Is detected by the logical sum operation unit 147.
And a logical sum operation of the detection result is performed for each pixel. Then, the result of the OR operation performed by the OR operation unit 147 is output as a line extraction result by the line extraction unit 31-0.

Next, the one-pixel horizontal line left detector 131 will be described with reference to FIG. 1-pixel horizontal line left detector 131
Detects a horizontal line having a width of one pixel as shown in FIG. 16F using a 3 × 3 pixel matrix. Now, 3 × 3 pixels are called _{A to I} as shown in FIG. 16A, and their pixel values are P _{A to} P _I. At this time, when all of the following equations (41) to (45) hold, the one-pixel horizontal line left detection unit 131 determines that there is a horizontal line of one pixel width as shown in FIG. 16 (a), a high level is output as a pixel value corresponding to D · E. Conversely, when even one of the equations (41) to (45) does not hold, the state shown in FIG. It is determined that there is no horizontal line having a width of one pixel, and a low level is output.

[0126] _{AVE (P D, P E)} -AVE (P A, P B)> TH41 ... (41) AVE (P D, P E) -AVE (P G, P H)> TH41 ... (42) MIN _{(P D, P E)>} TH42 ... (43) MAX (P A, P B) <TH43 ... (44) MAX (P G, P H) <TH43 ... (45) Note that, where, AVE (), MIN (), MAX
() Indicates the average value, the minimum value, and the maximum value of the pixel values in (), respectively. In addition, TH41 to TH44 are respectively predetermined thresholds.

By performing the above-described processing in the one-pixel horizontal line left detection unit 131, the horizontal line as shown in FIG. 16B can be detected, and the right end of the horizontal line can also be detected. Becomes

Next, the one-pixel horizontal line right detector 132 will be described with reference to FIG. 1 pixel horizontal line right detection unit 132
Detects a horizontal line of one pixel width as shown in FIG. 16 (j) using a 3 × 3 pixel matrix. When the following equations (46) to (50) are all satisfied, the one-pixel horizontal line right detection unit 132 determines that there is a one-pixel width horizontal line as shown in FIG. a) outputs a high level as a pixel value corresponding to EF in FIG.
When even one of the expressions 6) to (50) does not hold, FIG.
It is determined that there is no horizontal line having a width of one pixel as shown in (j), and a low level is output for each.

AVE (P _E , P _F ) -AVE (P _B , P _C )> TH44 ... (46) AVE (P _E , P _F ) -AVE (P _H , P _I )> TH44 ... (47) MIN (P _E , P _F )> TH 45 (48) MAX (P _B , P _C ) <TH 46 (49) MAX (P _H , P _I ) <TH 46 (50)

Next, the one-pixel vertical line detection section 133 will be described with reference to FIG. One pixel vertical line detection unit 133
Detects a vertical line having a width of one pixel as shown in FIG. 16 (g) using a 3 × 3 pixel matrix. 1 pixel vertical line detector 1
In 33, when all of the following equations (51) to (55) hold, it is determined that there is a vertical line of one pixel width as shown in FIG. 16 (g), and BE in FIG. Is output as a pixel value corresponding to the above, and conversely, when even one of the equations (51) to (55) does not hold, there is no vertical line of one pixel width as shown in FIG. And outputs a low level.

AVE (P _B , P _E ) −AVE (P _A , P _D )> TH47 (51) AVE (P _B , P _E ) −AVE (P _C , P _F )> TH47 (52) MIN _{(P B, P E)>} TH48 ... (53) MAX (P A, P D) <TH49 ... (54) MAX (P C, P F) <TH49 ... (55)

Next, the one-pixel vertical line under detection unit 134 will be described with reference to FIG. One pixel vertical line under detector 134
Detects a vertical line having a width of one pixel as shown in FIG. 16K using a 3 × 3 pixel matrix. 1 pixel vertical line detector 1
In 34, when all of the following equations (56) to (60) hold, it is determined that there is a vertical line of one pixel width as shown in FIG. 16 (k), and E · H in FIG. Is output as a pixel value corresponding to the above, and when at least one of the expressions (56) to (60) does not hold, there is no vertical line of one pixel width as shown in FIG. And outputs a low level.

AVE (P _E , P _H ) −AVE (P _D , P _G )> TH50 (56) AVE (P _E , P _H ) −AVE (P _F , P _I )> TH50 (57) MIN (P _E , P _H )> TH 51 (58) MAX (P _D , P _G ) <TH 52 (59) MAX (P _F , P _I ) <TH 52 (60)

Next, the one-pixel left oblique line detection section 135 will be described with reference to FIG. 1 pixel left oblique line detection unit 1
Reference numeral 35 detects a left-upward diagonal line having a width of one pixel as shown in FIG. 16H using a 3 × 3 pixel matrix. This one
When all of the following equations (61) to (65) hold, the pixel left diagonal on-detection unit 135 determines that there is a left-up diagonal line of one pixel width as shown in FIG.
A high level is output as a pixel value corresponding to A · E in (a). Conversely, if at least one of Equations (61) to (65) does not hold, 1 as shown in FIG. It is determined that there is no diagonal line of the pixel width rising to the left, and a low level is output for each.

AVE (P _A , P _E ) −AVE (P _D , P _G , P _H )> TH53 (61) AVE (P _A , P _E ) −AVE (P _B , P _C , P _F )> TH53 ... (62) MIN (P A, P E)> TH54 ... (63) MAX (P D, P G, P H) <TH55 ... (64) MAX (P B, P C, P F) <TH55 ... (65)

Next, the one-pixel left oblique line under detection unit 136 will be described with reference to FIG. 1 pixel left oblique line under detection unit 1
Reference numeral 36 uses a 3 × 3 pixel matrix to detect a left-upward diagonal line having a width of one pixel as shown in FIG. This one
When all of the following equations (66) to (70) hold, the pixel left oblique line under detection unit 136 determines that there is an obliquely upward diagonal line of one pixel width as shown in FIG.
A high level is output as a pixel value corresponding to E · I in (a). Conversely, when even one of the equations (66) to (70) does not hold, 1 as shown in FIG. It is determined that there is no diagonal line of the pixel width rising to the left, and a low level is output for each.

AVE (P _E , P _I ) −AVE (P _D , P _G , P _H )> TH56 (66) AVE (P _E , P _I ) −AVE (P _B , P _C , P _F )> TH56 ... (67) MIN (P _E , P _I )> TH 57 ... (68) MAX (P _D , P _G , P _H ) <TH58 ... (69) MAX (P _B , P _C , P _F ) <TH58 ... (70)

Next, the one-pixel right oblique line detector 137 will be described with reference to FIG. 1 pixel right oblique line detector 1
Reference numeral 37 uses a 3 × 3 pixel matrix to detect an oblique line rising to the right with a width of one pixel as shown in FIG. This one
When the following equations (71) to (75) are all satisfied, the pixel right oblique line detection unit 137 determines that there is a right-upward oblique line of one pixel width as shown in FIG.
A high level is output as a pixel value corresponding to CE in (a). Conversely, if even one of the equations (71) to (75) does not hold, 1 as shown in FIG. It is determined that there is no diagonal line with the pixel width rising to the right, and a low level is output for each.

AVE (P _C , P _E ) −AVE (P _A , P _B , P _D )> TH59 (71) AVE (P _C , P _E ) −AVE (P _F , P _H , P _I )> TH59 ... (72) MIN (P C, P E)> TH60 ... (73) MAX (P A, P B, P D) <TH61 ... (74) MAX (P F, P H, P I) <TH61 ... (75)

Next, the one-pixel right oblique line under detection unit 138 will be described with reference to FIG. 1 pixel right oblique line under detection unit 1
Numeral 38 detects a right-upward oblique line having a width of one pixel as shown in FIG. 16 (m) using a 3 × 3 pixel matrix. This one
When the following formulas (76) to (80) are all satisfied, the pixel lower right oblique line detection unit 138 determines that there is a right-upward oblique line of one pixel width as shown in FIG.
A high level is output as a pixel value corresponding to E · G in (a). Conversely, if at least one of Equations (76) to (80) does not hold, 1 as shown in FIG. It is determined that there is no diagonal line with the pixel width rising to the right, and a low level is output for each.

AVE (P _E , P _G ) −AVE (P _A , P _B , P _D )> TH62 (76) AVE (P _E , P _G ) −AVE (P _F , P _H , P _I )> TH62 ... (77) MIN (P E, P G)> TH63 ... (78) MAX (P A, P B, P D) <TH64 ... (79) MAX (P F, P H, P I) <TH64 ... (80)

Next, the two-pixel horizontal line left detector 139 will be described with reference to FIG. 2 pixel horizontal line left detector 13
No. 9 detects a horizontal line having a width of two pixels as shown in FIG. 21F using a 4 × 4 pixel matrix. Now, 4 × 4 pixels are called _{J to Y} as shown in FIG. 21A, and their pixel values are set to P _{J to} P _Y. When all of the following equations (81) to (85) hold, the two-pixel horizontal line left detection unit 139 determines that there is a horizontal line having a two-pixel width as shown in FIG. a) N ・ O ・ P ・ R ・
A high level is output as a pixel value corresponding to S · T. Conversely, when even one of the equations (81) to (85) does not hold, a horizontal line having a width of two pixels as shown in FIG. Is determined to be absent, and a low level is output for each.

AVE (P _N , P _O , P _P , P _R , P _S , P _T ) −AVE (P _J , P _K , P _L )> TH65 (81) AVE (P _N , P _O , P _P , P _R , P _S , P _T ) −AVE (P _V , P _W , P _X )> TH65 (82) MIN (P _N , P _O , P _P , P _R , P _S , P _T )> TH66 ... (83) MAX (P J, P K, P L) <TH67 ... (84) MAX (P V, P W, P X) <TH67 ... (85)

Next, the two-pixel horizontal line right detector 180 will be described with reference to FIG. 2 pixel horizontal line right detector 180
Detects a horizontal line having a width of two pixels as shown in FIG. 21 (j) using a 4 × 4 pixel matrix. When the following formulas (86) to (90) are all satisfied, the two-pixel horizontal line right detection unit 180 determines that there is a horizontal line having a two-pixel width as shown in FIG. a) O ・ P ・ Q ・ S ・
A high level is output as a pixel value corresponding to TU, and conversely, when even one of the equations (86) to (90) does not hold, a horizontal line having a width of two pixels as shown in FIG. Is determined to be absent, and a low level is output for each.

AVE (P _O , P _P , P _Q , P _S , P _T , P _U ) −AVE (P _K , P _L , P _M )> TH68 (86) AVE (P _O , P _P , P _Q , P _S , P _T , P _U ) −AVE (P _W , P _X , P _Y )> TH68 (87) MIN (P _O , P _P , P _Q , P _S , P _T , P _U )> TH69 (88) MAX (P _K , P _L , P _M ) <TH 70 (89) MAX (P _W , P _X , P _Y ) <TH 70 (90)

Next, the two-pixel vertical line detection unit 181 will be described with reference to FIG. 2-pixel vertical line detection unit 181
Detects a vertical line having a width of two pixels as shown in FIG. 21 (g) using a 4 × 4 pixel matrix. When the following formulas (91) to (95) are all satisfied, the two-pixel vertical-line horizontal protruding portion 181 determines that there is a vertical line having a two-pixel width as shown in FIG. (A) KLOPS
When a high value is output as a pixel value corresponding to T, and conversely, even if even one of Equations (91) to (95) does not hold,
It is determined that there is no vertical line having a width of two pixels as shown in FIG.

AVE (P _K , P _L , P _O , P _P , P _S , P _T ) −AVE (P _J , P _N , P _R )> TH 71 (91) AVE (P _K , P _L , P _O , P _P , P _S , P _T ) −AVE (P _M , P _Q , P _U )> TH 71 (92) MIN (P _K , P _L , P _O , P _P , P _S , P _T )> TH72 ... (93) MAX (P J, P N, P R) <TH73 ... (94) MAX (P M, P Q, P U) <TH73 ... (95)

Next, the two-pixel vertical line under detection unit 182 will be described with reference to FIG. 2-pixel vertical line under detector 182
Detects a vertical line having a width of two pixels as shown in FIG. 21K using a 4 × 4 pixel matrix. When all of the following equations (96) to (100) are satisfied, the two-pixel vertical line under detection unit 182 determines that there is a two-pixel width vertical line as shown in FIG. OPST of 21 (a)
A high level is output as a pixel value corresponding to W · X, and conversely, when at least one of Equations (96) to (100) does not hold, a vertical pixel having a width of two pixels as shown in FIG. It is determined that there are no lines, and each outputs a low level.

AVE (P _O , P _P , P _S , P _T , P _W , P _X ) −AVE (P _N , P _R , P _V )> TH74 (96) AVE (P _O , P _P , P _S , P _T , P _W , P _X ) −AVE (P _Q , P _U , P _Y )> TH74 (97) MIN (P _O , P _P , P _S , P _T , P _W , P _X )> TH75 ... (98) MAX (P N, P R, P V) <TH76 ... (99) MAX (P Q, P U, P Y) <TH76 ... (100)

Next, the two-pixel left oblique line detection unit 183 will be described with reference to FIG. 2 pixels left diagonal horizontal side 1
Reference numeral 83 denotes a two-pixel wide leftward oblique line as shown in FIG. 21H using a 4 × 4 pixel matrix. This 2
When all of the following equations (101) to (105) are satisfied, the pixel left oblique line detection unit 183 determines that there is an obliquely upward diagonal line having a width of two pixels as shown in FIG. a) outputs a high level as a pixel value corresponding to J, K, N, O, P, S, T, respectively.
When even one of the expressions (1) to (105) does not hold, FIG.
It is determined that there is no two-pixel-width diagonal line as shown in FIG.

[0151] _{AVE (P J, P K,} P N, P O, P P, P S, P T) -AVE (P R, P V, P W, P X)> TH77 ... (101) AVE (P _J , P _K , P _N , P _O , P _P , P _S , P _T ) −AVE (P _L , P _M , P _Q , P _U )> TH77 ... (102) MIN (P _J , P _K , P _N , P _O , P _P , P _S , P _T )> TH 78 (103) MAX (P _R , P _V , P _W , P _X ) <TH 79 (104) MAX (P _L , P _M , P _Q) , P _U ) <TH79 (105)

Next, the two-pixel left oblique line under detection unit 184 will be described with reference to FIG. 2-pixel left oblique line under detection unit 1
Numeral 84 detects an oblique line rising to the left with a width of two pixels as shown in FIG. 21 (l) using a 4 × 4 pixel matrix. This 2
When all of the following equations (106) to (110) hold, the pixel left oblique line under detection unit 184 determines that there is an obliquely upward diagonal line having a width of two pixels as shown in FIG. a) output high levels as pixel values corresponding to O, P, S, T, U, X, Y, respectively.
6) When even one of the expressions (110) does not hold, FIG.
It is determined that there is no left-upward diagonal line having a width of two pixels as shown in 1 (l), and outputs a low level, respectively.

[0153] _{AVE (P O, P P,} P S, P T, P U, P X, P Y) -AVE (P N, P R, P V, P W)> TH80 ... (106) AVE (P _O , P _P , P _S , P _T , P _U , P _X , P _Y ) −AVE (P _K , P _L , P _M , P _Q )> TH80 ... (107) MIN (P _O , P _P , P _S , P _T , P _U , P _X , P _Y )> TH 81 (108) MAX (P _N , P _R , P _V , P _W ) <TH 82 (109) MAX (P _K , P _L , P _M) , P _Q ) <TH82 (110)

Next, the two-pixel right oblique line detector 185 will be described with reference to FIG. 2-pixel right oblique line detector 1
Reference numeral 85 uses a 4 × 4 pixel matrix to detect a diagonally right-upward diagonal line having a width of two pixels as shown in FIG. This 2
When all of the following equations (111) to (115) are satisfied, the pixel right diagonal on-line detection unit 185 determines that there is a diagonally right diagonal line having a width of two pixels as shown in FIG. a) outputs a high level as a pixel value corresponding to L, M, O, P, Q, S, and T, and conversely, formula (11)
If any one of 1) to (115) does not hold, FIG.
It is determined that there is no upward slanting line having a width of two pixels as shown in 1 (i), and outputs a low level, respectively.

AVE (P _L , P _M , P _O , P _P , P _Q , P _S , P _T ) −AVE (P _J , P _K , P _N , P _R )> TH83 (111) AVE (P _L , P _M , P _O , P _P , P _Q , P _S , P _T ) −AVE (P _W , P _U , P _X , P _Y )> TH83 (112) MIN (P _L , P _M , P _O , P _P , P _Q , P _S , P _T )> TH 84 (113) MAX (P _J , P _K , P _N , P _R ) <TH 85 (114) MAX (P _W , P _U , P _X) , P _Y ) <TH85 (115)

Finally, the two-pixel lower right oblique line detector 186 will be described with reference to FIG. 2-pixel right oblique line under detection unit 1
Reference numeral 86 uses a 4 × 4 pixel matrix to detect a diagonally rightward diagonal line having a width of two pixels as shown in FIG. This 2
When all of the following expressions (116) to (120) are satisfied, the pixel right oblique line lower detection unit 186 determines that there is an upward diagonal line having a width of two pixels as shown in FIG. a) outputs a high level as a pixel value corresponding to O, P, R, S, T, V, W, and conversely, formula (11)
6) When even one of the expressions (120) does not hold, FIG.
It is determined that there is no diagonal line rising to the right with a width of two pixels as shown in FIG.

[0157] _{AVE (P O, P P,} P R, P S, P T, P V, P W) -AVE (P J, P K, P L, P N)> TH86 ... (116) AVE (P _O , P _P , P _R , P _S , P _T , P _V , P _W ) −AVE (P _Q , P _U , P _X , P _Y )> TH86 (117) MIN (P _O , P _P , P _{R, P S, P T,} P V, P W)> TH87 ... (118) MAX (P J, P K, P L, P N) <TH88 ... (119) MAX (P Q, P U, P X , P _Y ) <TH88 (120)

FIG. 27 is a block diagram showing an example of the configuration of the edge extraction unit 32. 27, the image data input to the edge extraction unit 32 is input to the 3 × 3 filter operation unit 151 and the floating binarization unit 153. The image data input to the 3 × 3 filter operation unit 151 is subjected to a filtering process with a preset 3 × 3 filter coefficient, and the operation result is output to the comparison unit 152. As the 3 × 3 filter coefficient, one that emphasizes an edge portion is desired, and as an example, a filter coefficient as shown in FIG. 28 is cited.

The operation result of the 3 × 3 filter operation unit 151 input to the comparison unit 152 is compared with a preset threshold value TH30, and the operation result of the 3 × 3 filter operation unit 151 is calculated based on the threshold value TH30. If it is larger, a high-level signal is output, and if it is equal or smaller, a low-level signal is output. The output signal from the comparison unit 152 is input to the AND operation unit 155.

On the other hand, the image data input to the floating binarizing section 153 is subjected to a floating binarizing process. Then, the processing result is input to the 5 × 5 exclusive OR operation unit 154. Details of the floating binarization unit 153 will be described later.

The image data input to the 5 × 5 exclusive OR operation unit 154 is subjected to an exclusive OR operation of 5 × 5 pixels centering on the target pixel. That is, 5 × 5 = 25
When all the pixels are high-level signals or low-level signals, a low-level signal is obtained as a calculation result, and when even one pixel is different, a high-level signal is obtained. The operation result of the 5 × 5 exclusive OR operation unit 154 is input to the AND operation unit 155. Then, in the OR operation unit 155, the output signal from the comparison unit 152 and the output signal from the AND operation unit 155 are AND-operated for each pixel, and the operation result is output as the output result of the edge extraction unit 32. You.

The details of the floating binarization section 153 will be described with reference to FIG. The image data input to the floating binarization unit 153 is input to the comparison unit 159 and the 3 × 3 average value calculation unit 156. The 3 × 3 average value calculation unit 156 calculates the average value of the 3 × 3 pixel values centered on the target pixel, and outputs the result to the addition unit 157. Adder 1
The calculation result of the 3 × 3 average value input to 57 is added to a predetermined addition value Val, and the result is output to the limit processing unit 158.

In the limit processing section 158, upper and lower limit values are set in advance.
7 is limited to the range of the upper and lower limit values. That is, the upper limit value is set when the calculation result input from the adding unit 157 is larger than the upper limit value, the lower limit value is set when the calculated result is smaller than the lower limit value, and the adding unit 157 is used when neither is the case. Is output as the calculation result of the limit processing unit 158.

The calculation result output from the limit processing unit 158 is input to a comparison unit 159, and the comparison operation with the image data input to the comparison unit 159, that is, the image data input to the floating binarization unit 153, is performed. Done. When the image data input to the floating binarization unit 153 is larger than the operation result of the comparison unit 159, a high-level signal is output, and when the image data is equal or smaller, a low-level signal is output. The output signal from the comparison unit 152 is the 5 × 5 exclusive OR operation unit 1
Input to 54.

As the processing in the edge extracting section 32, a known technique other than the above-described processing may be used. However, it is desirable that the processing of extracting the outline of a line segment or the like be as accurate as possible.

The details of the first to N-th enlargement units 33-1 to 33-N will be described with reference to FIGS. 29 and 30. In addition, the 1st expansion part 33-1-the Nth expansion part 3
Since 3-N performs almost the same processing, the first enlargement unit 33-1 will be described here as an example. FIG. 29 is a diagram for explaining the processing content in the first enlargement unit 33-1. FIG. 30 is a diagram for explaining the enlargement condition determination in the first enlargement unit 33-1.

The first magnifying unit 33-1 receives the image data from which the line has been extracted by the line extracting unit 33-1 and the image data from which the edge has been extracted by the edge extracting unit 32. The image data output from the line extracting unit 31-1 is:
Since the original image data, that is, the image data input to the line segment extraction unit 12 has been reduced by １ ／ each in the vertical and horizontal directions, the input line extraction unit 3
In step 1-1, the image data subjected to the line extraction is enlarged twice vertically and horizontally.

However, if the enlargement process is simply performed on the image data from which the line has been extracted by the line extracting unit 31-1, the result will be a step-like rattling. This will be described with reference to FIG. 29. When an image as shown in FIG. 29E is input to the line segment extraction unit 12, the line extraction unit 31-1 obtains an extraction result as shown in FIG. Can be When this process is simply performed to enlarge the image twice vertically and horizontally, FIG.
Although the result is as shown in FIG. 9 (c), the result is that the step-like rattling is larger than that of FIG. 29 (e), and the line segment faithful to the input image is not extracted.

Therefore, the first enlarging unit 33-1 refers to the image data from which the edge has been extracted by the edge extracting unit 32 while referring to the image data from which the line has been extracted by the line extracting unit 31-1. Perform enlargement processing. At this time, the output from the first enlargement unit 33-1 is determined by the condition determination as shown in FIG. For example, when a certain pixel of the image data input from the line extraction unit 31-1 is at a low level (LOW), the output from the first enlargement unit 33-1 is at a low level for all 2 × 2 pixels.

On the other hand, when a certain pixel of the image data input from the line extraction house 31-1 is at a high level (HIGH),
Condition determination is performed based on the values of the peripheral pixels of the pixel and the input value from the edge extraction unit 32, and the output from the first enlargement unit 33-1 is determined. By performing the enlarging process of the image data subjected to the line extraction by the line extracting unit 31-1 according to the above condition determination, the input image data from the line extracting unit 31-1 as shown in FIG. , Based on the input image data from the edge extraction unit 32 as shown in FIG.
Image data subjected to the enlargement processing as shown in FIG. 29D is obtained.

The details of the line segment correction section 17 will be described with reference to the flowchart of FIG. The line segment correction unit 17 extracts the line segment extraction image input from the line segment extraction unit 12 from the labeling image input from the labeling unit 13 and the attribute determination result for each label input from the first attribute determination unit 16. Correct on a pixel-by-pixel basis.

Referring to the flowchart of FIG. 31, first, the label number assigned to the target pixel is checked from the labeling image (step S401). Then, it is determined whether or not the label number exists (step S402). If the label number has been assigned, the attribute of the label number is checked from the attribute determination result for each label (step S403), and the attribute is determined (step S404). . If the attribute is a photograph / picture area, the extraction result is corrected even when the pixel of interest is extracted by the line segment extraction unit 12. That is, it is determined that the target pixel is not a line segment, and the target pixel is set to a low level (LOW) as a correction result of the line segment correction unit 17 (step S405).

On the other hand, if it is determined in step S402 that no label number has been assigned to the target pixel, or if it is determined in step S404 that the attribute of the pixel of interest is a uniform color area, the line segment extraction unit 12 The result is used (step S406). That is, when the line segment extraction result is low level, the low level is determined, and when the line extraction result is high level, the character / line segment written in the uniform color area is determined. Set to low level.

After the completion of the processing in step S405 or S406, since the correction of one pixel has been completed, the pixel of interest is moved to the next pixel (step S407). Then
It is determined whether or not the correction processing has been completed for all the pixels (step S408). When the correction processing has been completed, the processing in the line segment correction unit 17 is completed, and there is a pixel for which the correction processing has not been completed. Then, the process returns to step SS401 to repeat the above-described processing. The image output from the line segment correction unit 17 is defined as an image area separation image output from the image area separation unit 9.

As described above, according to the image area separation unit 9 according to the first embodiment of the present invention, labeling is performed on a binarized image, and a feature amount is calculated for each label area. By adopting a configuration in which the attribute of the label area is determined from the feature amount and the line segment extracted from the input image is corrected, characters that are mixed with characters, line drawings, photographs, halftone dots, etc. A high-precision image area separation process that extracts a line drawing and does not cause erroneous extraction in a photograph / pattern region can be performed, so that a good reproduced image can be obtained.

Next, an image area separating section 9 according to a second embodiment of the present invention will be described with reference to the drawings. In the following description, the same reference numerals are given to the processing units having the same processing contents as those of the first embodiment, and the description thereof will be omitted.

FIG. 32 is a block diagram showing the structure of the image area separating section 9 according to the second embodiment of the present invention. In FIG. 32, a binarizing unit 11 reads out image data (resolution of 400 dpi, L ^* a ^* b ^* image signal of 8 bits for each pixel) stored in the data storage unit 2, and performs binarization processing. Hereinafter, this image data is referred to as an L ^* a ^* b ^* original image. The L ^* a ^* b ^* original image is also read out from a line segment extraction unit 12, label feature amount calculation units 14-1 and 14-2, and an edge extraction unit 61, which will be described later.

The binarizing section 11 converts the read image data (L ^* a ^* b ^* original image) into a foreground area, for example, a character / line segment on a white background, a colored area, or a photograph / picture An area or the like is extracted, and an extracted pixel is set to a high level, and an unextracted pixel is set to a low level. An image signal of 1 bit for each pixel is output. Since the details of the binarization unit 11 have already been described, the description here is omitted.

Taking FIGS. 33 to 35 as an example, FIG.
Assuming that the image of (a) is input to the binarization unit 11, a binarized image is output as shown in FIG. 2
The binary image signal output from the binarizing unit 11 is input to the labeling unit 13-1, and one label is assigned to each of the connected extracted pixel regions.

A known technique may be used as a labeling method. Referring to FIG. 33, the binarized image of FIG. 33B is input to the labeling unit 13-1, and labels from label number P1 to label number Pn are given as shown in FIG. Hereinafter, an image to which a label is added for each connected region is referred to as a labeling image, a labeling image obtained by the labeling unit 13-1 is hereinafter referred to as a parent labeling image, and each label region is referred to as a parent label region. Shall be called.

The parent labeling image output from the labeling unit 13-1 is input to the label feature amount calculation unit 14-1 and the line segment correction unit 17. The label feature calculation unit 14-1 calculates a feature amount for each parent label area from the parent labeling image output from the labeling unit 13-1 and the L ^* a ^* b ^* original image.

Referring to FIG. 33, the parent labeling image to which the label shown in FIG.
4-1 is input, and the L ^* a ^* b ^* original image data of FIG. 33A is also referred to to calculate the feature amount for each parent label. In the present embodiment, as the feature amount, the number of applicable label-assigned pixels, the circumscribed rectangle position, the circumscribed rectangle size, the pixel density (= the number of applicable label-assigned pixels / the circumscribed rectangular area), the average value, the variance value, and the class division value frequency Although the class division and the like are calculated, other feature amounts may be calculated and used for the determination process in the second attribute determination unit 65 described later.

The details of the label feature calculator 14-1 are the same as those of the label feature calculator 14, so that the description thereof will be omitted. The feature amount for each parent label calculated by the label feature calculation unit 14-1 is stored in the small area removal unit 15-1 together with the parent labeling image to which the label is added by the labeling unit 13-1.
Is compared with a predetermined threshold value, and a parent label having a region size equal to or smaller than the threshold value is removed.

In the case of FIG. 33, assuming that the threshold value is, for example, circumscribed rectangle width = circumscribed rectangle height = 80, the areas of label numbers P4 to Pn are removed from the parent labeling image of FIG. A parent labeling image as shown in (d) is obtained. That is, by determining the area estimated to be a character from the circumscribed rectangle size and removing the area in advance, it is possible to reduce the processing in the various processing units in the subsequent stages. The parent labeling image from which the small area has been removed by the small area extracting section 15-1 is input to an edge / line segment removing section 63 and an inclusion relation determining section 64, which will be described later.

On the other hand, the edge extraction unit 61 determines the L ^* a ^* b ^{* of the} L ^* a ^* b ^* original image data stored in the data storage unit 2.
^* Read the original image and perform edge extraction processing. The edge extraction processing in the edge extraction unit 61 may be a known technique. For example, a 3 × 3 pixel size filtering processing is performed around the pixel of interest, and the calculation result is compared with a predetermined threshold value. If they are equal, it is determined to be an edge, and a high-level signal is output. If less than the threshold, it is determined to be non-edge and a low-level signal is output.

The 3 × 3 filter coefficients include, for example, a coefficient having a center of 8 and a periphery of −1. Referring to FIGS. 33 and 34, FIG.
34 (a) is obtained from the original image of FIG. The edge extraction image output from the edge extraction unit 61 is input to the image synthesis unit 62.

[0187] The line segment extracting section 12 is also provided in the data storage section 2.
Load the L ^* original image among the stored L ^* a ^* b ^* original image, performs the line segment extraction processing. Details of the line segment extraction unit 12 have already been described, and thus description thereof will be omitted. However, referring to FIGS. 33 and 34, a line segment extraction image of FIG. 34B is obtained from the original image of FIG. Line segment extraction unit 12
Are also input to the image synthesizing unit 62.

The image synthesizing unit 62 receives the edge extracted image and the line segment extracted image from the edge extracting unit 61 and the line segment extracting unit 12, calculates the logical sum of the pixels at the same coordinate position, and executes the expansion processing unit. It is output to 66. The dilation processing unit 66 performs dilation processing of the logical sum image of the edge extracted image and the line segment extracted image input from the image synthesizing unit 63 in an N × N size centering on the target pixel. That is, if at least one high-level pixel exists in the N × N range around the target pixel, a high-level signal is output. Output as a signal.

The reason why the expansion processing is performed here is that an edge / line segment is removed by an edge / line segment removing unit 63 described later.
Although labeling and feature extraction are performed on the image, the peripheral pixels of the edge / line are easily affected by the edge / line at the time of image input, and it is difficult to obtain an accurate feature. Therefore, by expanding the edge / line segment extraction region in advance and removing the expanded region, it is possible to obtain an accurate feature amount that is not affected by the edge / line segment. Therefore, the size N at which the expansion processing is performed by the expansion processing unit 66 is:
The size may be such that the effects of edges and line segments can be removed, for example, N = 5.

Referring to FIG. 34, an image shown in FIG. 34 (c) is obtained by combining and expanding the edge extracted image shown in FIG. 34 (a) and the line segment extracted image shown in FIG. 34 (b). The edge / line segment removing unit 63 includes a small area removing unit 15-
1, the parent labeling image from which the small area has been removed, and the expanded image of the combined edge / line segment image output from the dilation processing unit 66 are input.
High level pixels of the dilated image of the line segment composite image are removed.

That is, when the target pixel of the expanded image of the edge / line combined image is at a high level, the label number at the same coordinate position is removed from the parent rappelling image, and no label is assigned. Referring to FIG. 33 and FIG. 34, from the parent labeling image from which the small area has been removed in FIG.
The combined edge / line image shown in FIG. 34C is removed, and FIG.
An edge / line segment removal rappelling image of 4 (d) is obtained. The parent labeling image from which the edges and line segments have been removed is input to the labeling unit 13-2.

In the labeling section 13-2, the parent labeling image from which the small area has been removed by the small area removing section 15-1 and the edge / line segment has been removed by the edge / line segment removing section 63 is The labeling process is performed again so that the same label is assigned to the connected regions. A well-known technique may be used for the labeling process. The labeling image obtained here is hereinafter referred to as a child labeling image.

Referring to FIG. 34 and FIG. 35, FIG.
Labeling is again performed on the image from which the small area and the edge / line segment have been removed in (d), and a child labeling image to which labels from label number C1 to label number Cm are given in FIG. 35A is obtained. Although FIG. 35A also shows small label areas other than the label numbers C1 to Cm, for example, two circular areas surrounded by a line of the character “8”, FIG. In the figure, label numbers are omitted.

The child labeling image labeled by the labeling unit 13-2 is output to the label feature amount calculation unit 14-.
2 is input. The label feature amount calculation unit 14-2 receives the child labeling image from the labeling unit 13-2, reads the L ^* a ^* b ^* original image data from the image storage unit 2, and calculates the feature amount for each child label. Is done. The details of the label feature calculation unit 14-2 are also the same as those of the label feature amount calculation unit 14, and a description thereof will be omitted.

The feature amount of each child label area calculated by the label feature calculation section 14-2 is input to the small area removing section 15-2 together with the child labeling image to which the label is added by the labeling section 13-2. The region size calculated for each label is compared with a predetermined threshold, and child labels having a region size equal to or smaller than the threshold are removed. In the case of FIG. 35 (a), the minute area in which the description of the label number is omitted corresponds to the area to be removed by the small area removing unit 15-2. In addition, the method of small area removal is
It may be the same as that of the small area removing unit 15-1 described above, but it is desirable that the threshold value for determining a small area is smaller than that of the small area removing unit 15-1.

The child labeling image from which the small areas have been removed is input to the inclusion relation determining unit 64 and the first attribute determining unit 16. The inclusion relation determining unit 64 includes a small area removing unit 15-1.
The parent labeling image from which the small region has been removed and the child labeling image from which the small region has been removed by the small region removing unit 15-2 are input, and their inclusion relation, that is, which parent label region is divided and which child label is It is determined whether an area has been generated.

Although not shown, the label feature amount calculated by the label feature amount calculation units 14-2 and 14-2 is input to the inclusion relationship determination unit 64 to improve the efficiency of the inclusion relationship determination process. Is also possible. The inclusion relationship between the parent label region and the child label region determined by the inclusion relationship determination unit 64 is input to the second attribute determination unit 65. Details of the inclusion relation determination unit 64 will be described later.

On the other hand, the feature amount for each child label area calculated by the label feature calculation section 14-2 is input to the first attribute determination section 16, and the attribute for each child label area is determined using the calculated feature amount. judge. In the present embodiment, the number of pixels having the corresponding label, the variance value, the frequency of the class division value, and the class variance are used as the feature amounts, and it is determined whether the attribute is a photograph / picture region or a uniform color region.

Since the details of the first attribute judging section 16 have already been described, a description thereof will be omitted here.
The label feature calculation unit 14-2 calculates the feature amount for each child label area, and the first attribute determination unit 16 determines the attribute,
For example, if there is a table or the like in which each cell has a different color, the cells are separated by ruled lines.
One is a child label area, and the feature amount calculation / attribute determination is performed on a cell-by-cell basis, thus enabling highly accurate attribute determination.

Even in a table having no ruled line, a boundary between cells can be extracted by the edge extracting unit 61.
Similarly, each cell becomes a child label area, and it becomes possible to perform feature amount calculation and attribute determination in cell units. First
The attribute determination result for the child label determined by the attribute determination unit 16 is input to the second attribute determination unit 65.

The second attribute judging section 65 checks the inclusion relation between the parent label area and the child label area inputted from the inclusion relation judging section 64, the attribute judgment result for the child label area inputted from the first attribute judging section 16, and Is used to determine the attribute of the parent label area. In the present embodiment, it is determined whether the attribute is a photograph / pattern area or a uniform color area. The attribute determination result for the parent label in the second attribute determination section 65 is input to the line segment correction section 17. The details of the second attribute determination unit 65 will be described later.

In the line segment correction section 17, the labeling section 13-
From the parent rappelling image input from No. 1 and the attribute determination result for each parent label input from the second attribute determination unit 65,
The line segment extraction image input from the line segment extraction unit 12 is corrected for each pixel. Since the details of the line segment correction unit 17 have been described above, a detailed description thereof will be omitted. However, in brief, line segments included in the parent label area determined to be a photograph / picture area are deleted, and other lines are deleted. Leave the minutes alone. The image output from the line correction unit 17 is stored in the image storage unit 2 as an image output from the image area separation unit 9.

The details of the inclusion relation judging section 64 are shown in FIG.
This will be described with reference to FIGS. The parent labeling image output from the small area removing section 15-1 and the child labeling image output from the small area removing section 15-2 are input to the inclusion relation determining section 64, and the parent label area and the child label area are output. An inclusion relationship is determined.

That is, when all the pixels of the child label area are included in the parent label area, it is determined that the parent label area and the child label area have an inclusive relation. In the case of FIG. 33D and FIG. 35A, it is determined that the child label area of the child label number C1 is included in the parent label area of the parent label number P1. Similarly, the child label areas of the child label numbers C14 to Cm are included in the parent label area of the parent label number P2, and the child label areas of the child label numbers C2 to C13 are included in the parent label area of the parent label number P3. Is determined, and a determination result table as shown in FIG. 36 is created. Note that FIG.
Null in 6 indicates that there is no further child label area included.

[0205] The inclusion relation determination result determined by the inclusion relation determination section 64 is output to the second attribute determination section 65. Details of the second attribute determination section 65 will be described with reference to FIG. The second attribute determination unit 65 receives the inclusion relationship determination result from the inclusion relationship determination unit 64 and the child label region attribute determination result from the first attribute determination unit 16 and determines the attribute of the parent label region.

The second attribute judging section 65 looks at the attributes of a plurality of child label areas included in the parent label area, and if the number of child label areas judged to be photo / picture areas is equal to or greater than a predetermined threshold value, The label area is determined to be a photograph / picture area, and if less than the threshold, it is determined to be a uniform color area. However, when the parent label area includes one child label area, the attribute determination result of the parent label area is the same as the attribute determination result of the child label area. In addition, the parent label area has one child label area.
If none is included, the parent label area is determined to be a uniform color area.

In the case of FIG. 37, when the threshold value is set to 2, the parent label area having the parent label number P1 has the child label number C1.
Since only one child label area is included and is a uniform color area, the parent label area of the parent label number P1 is determined to be a uniform color area. Further, since the parent label area having the parent label number P2 includes the child label areas having the child label numbers C14 to Cm and the number of the photograph / picture areas is two, the parent label area having the parent label number P2 is photographed. -It is determined as a picture area. In addition, since the child label number includes the child label areas of C2 to C13 and the number of the photograph / picture areas is one, the parent label area of the parent label number P3 is determined to be the uniform color area.

In the above description, the attribute of the parent label area is determined by comparing the number of child label areas of the photo / picture area with the threshold value. There is also a method of comparing the ratio of the number of label areas with a threshold. The attribute determination result of the parent label area output from the second attribute determination section 65 is input to the line segment correction section 17.

As described above, according to the image area separating unit 9 according to the second embodiment of the present invention, a labeling is performed on the binarized image to generate a parent labeling image, and a small labeling image is generated from the parent labeling image. Labeling is performed again on the image from which the region, edge, and line segment have been removed to create a child labeling image, and a feature value is calculated for each child label region, and the attribute of the region is determined from the feature value. In addition, by adopting a configuration in which the attribute of the parent label is determined from the inclusion relationship between the parent label area and the child label area and the line segment extracted from the input image is corrected, characters, line drawings, photographs, halftone dots, etc. are mixed. A good reproduction image can be obtained because the image and the line image can be accurately extracted from the image and a high-precision image area separation process can be performed without erroneous extraction in the photograph / pattern area. It becomes possible.

In the second embodiment, an edge extracting section 61 is provided in addition to the line segment extracting section 12, and the line segments and edges extracted by the extracting sections 12 and 61 are combined with the image synthesizing section 62.
And the combined line segment and edge are removed from the binarized image data by the edge / line segment removing unit 63. However, the image combining unit 62 does not combine the line segment and the edge. It is also possible to adopt a configuration in which only the edge is removed from the binarized image data, or a configuration in which only the line segment is removed from the binarized image data without providing the edge extraction unit 61.

[0211]

As described above, according to the present invention,
When performing an appropriate reproduction process for each area on an image that contains characters, line drawings, photographs, halftone dots, etc., and outputting the image,
In addition to the white background, even in a uniform color background, characters and line drawings can be faithfully extracted from an input image, and high-precision image area separation processing can be performed without erroneous extraction in a photograph / pattern area. Therefore, a high-quality image can be obtained.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a basic configuration of an image processing apparatus according to the present invention.

FIG. 2 is a block diagram illustrating a configuration of an image area separation unit according to the first embodiment of the present invention.

FIG. 3 is a diagram (part 1) illustrating an image change in a processing process in the image area separation unit according to the first embodiment.

FIG. 4 is a diagram (part 2) illustrating an image change in the process of processing in the image area separation unit according to the first embodiment.

FIG. 5 is a diagram showing a feature amount for each label.

FIG. 6 is a block diagram illustrating an example of a configuration of a binarization unit.

FIG. 7 is a diagram illustrating an image change in a binarization unit.

FIG. 8 is a block diagram illustrating an example of a configuration of a shadow extraction unit.

FIG. 9 is a block diagram illustrating another example of the configuration of the shadow extraction unit.

FIG. 10 is a flowchart illustrating a flow of a process according to another example of the binarization unit.

FIG. 11 is a diagram showing a histogram of L ^* image data.

FIG. 12 is a flowchart illustrating a flow of a process in a label feature amount calculation unit.

FIG. 13 is a flowchart illustrating a flow of processing in a first attribute unit.

FIG. 14 is a block diagram illustrating an example of a configuration of a line segment extraction unit.

FIG. 15 is a block diagram illustrating an example of a configuration of a line extracting unit.

FIG. 16 is a diagram illustrating the types of lines detected by a line extraction unit having a width of one pixel.

FIG. 17 is a block diagram illustrating an example of a configuration of a one-pixel horizontal line detection unit.

FIG. 18 is a block diagram illustrating an example of a configuration of a one-pixel vertical line detection unit.

FIG. 19 is a block diagram illustrating an example of a configuration of a one-screen left diagonal line detection unit.

FIG. 20 is a block diagram illustrating an example of a configuration of a one-screen right diagonal line detection unit.

FIG. 21 is a diagram illustrating types of lines detected by a line extraction unit having a width of two pixels.

FIG. 22 is a block diagram illustrating an example of a configuration of a two-pixel horizontal line detection unit.

FIG. 23 is a block diagram illustrating an example of a configuration of a two-pixel vertical line detection unit.

FIG. 24 is a block diagram illustrating an example of a configuration of a two-screen left diagonal line detection unit.

FIG. 25 is a block diagram illustrating an example of a configuration of a two-screen right diagonal line detection unit.

FIG. 26 is a block diagram showing another example of the configuration of the line extracting unit.

FIG. 27 is a block diagram illustrating an example of a configuration of an edge extraction unit.

FIG. 28 is a diagram illustrating an example of a filter coefficient.

FIG. 29 is a diagram for explaining processing contents in a first enlargement unit.

FIG. 30 is a diagram for explaining enlargement condition determination in a first enlargement unit.

FIG. 31 is a flowchart showing the flow of processing in a line segment correction unit.

FIG. 32 is a block diagram illustrating a configuration of an image area separation unit according to a second embodiment of the present invention.

FIG. 33 is a diagram (part 1) illustrating an image change in a process in the image area separation unit according to the second embodiment.

FIG. 34 is a diagram (part 2) illustrating an image change in a process of processing in the image area separation unit according to the second embodiment.

FIG. 35 is a diagram (part 3) illustrating an image change in a processing process in the image area separation unit according to the second embodiment.

FIG. 36 is a diagram illustrating an example of a relationship between a parent label area and a child label area in an inclusion relation determination unit.

FIG. 37 is a diagram illustrating determination of a parent label area attribute from a child label area attribute of a second attribute part.

FIG. 38 is a block diagram illustrating an example of a configuration of an image area separation unit according to the related art.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Image input part, 2 ... Data storage part, 3 ... Operation processing part,
5: gradation correction section, 6: color signal conversion section, 8: image output section,
9: image area separation unit, 11: binarization unit, 12: line segment extraction unit
13, 13-1, 13-2 ... labeling part, 14, 14
-1, 14-2... Label feature amount calculation unit, 15, 1-1,
15-2: small area extraction unit, 16: first attribute determination unit, 17
... Line segment corrector, 61 edge extractor, 62 image synthesizer, 63 edge / line segment remover, 64 inclusion relation determiner, 65 second attribute determiner

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 5B057 CA08 CA12 CA16 CB07 CB08 CB12 CB16 CE02 CE03 CE08 CE09 CH18 DA08 DB02 DB09 DC05 DC16 DC36 5C076 AA02 AA11 AA19 AA36 BA03 BA06 5C077 LL20 MP02 MP05 MP06 MP07 NP19 PP03 PP03 PP05 PP23 PP27 PP28 PP43 PP46 PP47 PP58 PP61 PP68 PQ08 PQ12 PQ19 PQ20 RR02 5L096 AA02 AA06 BA12 CA24 EA35 FA03 FA06 FA33 FA42 FA43 FA44 FA45 FA78 GA34

Claims (31)

[Claims] An input unit for inputting image data; a binarizing unit for performing a process of separating a foreground and a background on the image data input by the input unit; and an image data input by the input unit. A line segment extracting unit that extracts a line segment having a width of a predetermined number of pixels, a first labeling unit that assigns the same label to connected pixels that have been foreground by the binarizing unit, For each label area to which the same label is assigned by the first labeling means, a first label feature quantity calculating means for calculating an image feature quantity of the label area, and for each label area calculated by the first label feature quantity calculating means First attribute determining means for determining an attribute of the label area from the label feature amount of the first and second label attributes, and the first attribute for each label area labeled by the first labeling means An image processing apparatus comprising: a line segment correcting unit that corrects a line segment extracted by the line segment extracting unit using an attribute determination result by a determining unit. 2. A line segment removing unit for removing a line segment extracted by the line segment extracting unit from the image data binarized by the binarizing unit; and a foreground set by the binarizing unit. A second labeling unit for assigning the same label to pixels and connected pixels from which a line segment has been removed by the line segment removing unit; a label region labeled by the first labeling unit and the second labeling unit Relationship judgment means for judging the inclusion relation of the label areas labeled by the labeling means, and the attribute judgment result by the first attribute judgment means for the label area labeled by the second labeling means and the inclusion by the inclusion relation judgment means From the judgment result,
A second attribute determining unit that determines an attribute of the label region labeled by the first labeling unit, wherein the line segment correcting unit is configured to perform the second attribute determination for each of the label regions labeled by the first labeling unit. 2. The image processing apparatus according to claim 1, wherein the line segment extracted by the line segment extracting unit is corrected using an attribute determination result by the attribute determining unit. 3. An image processing apparatus comprising: an edge extracting unit configured to extract an edge from image data input by the input unit; and an image data binarized by the binarizing unit.
An edge removing unit that removes an edge extracted by the edge extracting unit; and an identical label for a pixel that has been foregrounded by the binarizing unit and that is connected to a pixel whose edge has been removed by the edge removing unit. A second labeling unit to be provided; an inclusion relationship determining unit that determines an inclusion relationship between a label region labeled by the first labeling unit and a label region labeled by the second labeling unit; and the second labeling unit. From the attribute determination result by the first attribute determination unit and the inclusion determination result by the inclusion relationship determination unit with respect to the label area labeled by
A second attribute determining unit that determines an attribute of the label region labeled by the first labeling unit, wherein the line segment correcting unit is configured to perform the second attribute determination for each of the label regions labeled by the first labeling unit. 2. The image processing apparatus according to claim 1, wherein the line segment extracted by the line segment extracting unit is corrected using an attribute determination result by the attribute determining unit. 4. An edge extraction unit for extracting an edge from the image data input by the input unit, and a line segment extracted by the line segment extraction unit and an edge extracted by the edge extraction unit are combined. From the image data binarized by the binarizing unit,
An edge / line segment removing unit for removing an edge and a line segment synthesized by the image synthesizing unit; and a pixel set as a foreground by the binarizing unit and an edge and a line segment by the edge / line segment removing unit. Second to give the same label to the removed connected pixels
Labeling means; inclusion relation determining means for determining the inclusion relation of each of the label area labeled by the first labeling means and the label area labeled by the second labeling means; and labeling by the second labeling means. From the attribute determination result for the label area by the first attribute determination unit and the inclusion determination result by the inclusion relationship determination unit,
Second attribute determining means for determining an attribute of a label area labeled by the first labeling means, wherein the component correcting means comprises a second attribute for each label area labeled by the first labeling means. 2. The image processing apparatus according to claim 1, wherein the line segment extracted by the line segment extracting unit is corrected using an attribute determination result by the determining unit. 5. The apparatus according to claim 1, further comprising: a first small area removing unit configured to remove a label area having a predetermined size or less from a label feature amount for each label area calculated by the first label feature amount calculating unit. The image processing apparatus according to claim 1, wherein the determining unit determines an attribute of the label area that has not been removed by the first small area removing unit. 6. The first small area removing means compares the label area size calculated by the first label feature quantity calculating means with a predetermined size, and removes an area whose label area size is equal to or smaller than a predetermined size. The image processing apparatus according to claim 5, wherein: 7. A second label feature value calculating means for calculating an image feature value for each area labeled by the first labeling means, and a label area size calculated by the second label feature value calculating means being a predetermined size. 7. The image processing apparatus according to claim 5, further comprising: a second small area removing unit that removes an area whose label area size is equal to or smaller than a predetermined size, as compared with. 8. The apparatus according to claim 7, wherein the area size to be removed by said second small area removing means is equal to or smaller than the area size to be removed by said first small area removing means. Image processing device. 9. An expansion processing means for expanding an extraction area obtained by synthesis by the image synthesis means by a predetermined size, wherein the edge / line segment removing means is expanded by the expansion processing means. The image processing apparatus according to claim 4, wherein the extracted region is removed. 10. The method according to claim 1, wherein the first label feature amount calculating means calculates at least a class division value and a class variance ratio of the pixel value of the label area. Image processing device. 11. The image processing apparatus according to claim 1, wherein said first label feature amount calculating means calculates at least a variance of a pixel value of a label area. 12. The image processing apparatus according to claim 1, wherein said first label feature amount calculating means calculates at least a size of a label area. 13. The image processing apparatus according to claim 7, wherein said second label feature amount calculating means calculates at least a size of a label area. 14. The image processing apparatus according to claim 1, wherein the first attribute determination unit includes at least a photograph / picture area and a uniform color area as a determination result. 15. The image processing apparatus according to claim 2, wherein the second attribute determination unit includes at least a photograph / pattern area and a uniform color area as a determination result. 16. The method according to claim 1, wherein the first attribute determination unit uses a variance of the label area to be determined to determine whether the label area to be determined is a photograph
The image processing apparatus according to claim 1, wherein the image processing apparatus determines whether the area is a picture area or a uniform color area. 17. The method according to claim 1, wherein the first attribute determining unit uses a class division value and a class distribution ratio of the label region to be determined.
17. The method according to claim 1, wherein the determination target label area is a photograph / picture area or a uniform color area.
The image processing device according to any one of the above. 18. The method according to claim 18, wherein the inclusion relation determining means determines that there is inclusion when the area labeled by the second labeling means is included in the area labeled by the first labeling means. An image processing apparatus according to claim 2. 19. The method according to claim 19, wherein, if the area determined to be included by the inclusion relation determining means is a photo / pattern label area if the area is equal to or greater than a predetermined threshold, the second attribute determining means also determines that the area to be included is a photo. 19. The image processing apparatus according to claim 2, wherein the image processing apparatus determines that the area is a picture label area. 20. The second attribute judging unit, when the region judged to be included by the inclusion relation judging unit is a photo / pattern label region in a region having a ratio equal to or greater than a predetermined threshold value. The image processing apparatus according to any one of claims 2 to 19, wherein it is determined that the image is also a photo / picture label area. 21. The method according to claim 7, wherein the first attribute determining unit determines an attribute of the label area that has not been removed by the second small area removing unit. Image processing device. 22. The binarizing means, wherein shadow extracting means for extracting a high-density area from the image data input by the input means, and halftone extracting means for extracting an area composed of halftone dots And a logical sum operation means for performing a logical sum operation on each of the extraction results of the shadow extraction means and the halftone dot extraction means for each pixel.
The image processing device according to claim 21. 23. A binarizing unit, comprising: a histogram creating unit that creates a histogram of pixel values for the image data input by the input unit; and a binarizing threshold based on the histogram created by the histogram creating unit. And a binarization threshold value determining unit that determines the threshold value, and using the binarization threshold value determined by the binarization threshold value determination unit,
22. The image processing apparatus according to claim 1, wherein binarization of the image data input by the input unit is performed. 24. The line segment correcting unit corrects a line segment extraction result by the line segment extracting unit when the attribute of the target pixel is a photograph / pattern region, and corrects the result when the attribute of the target pixel is a uniform color region. The image processing apparatus according to any one of claims 1 to 23, wherein a line segment extraction result by the line segment extraction unit is used as a correction result. 25. A binarization step of performing a process of separating a foreground and a background from input image data, and a line segment extracting a line segment having a width of a predetermined number of pixels from the input image data. An extracting step, a first labeling step of assigning the same label to the connected pixels that have been set to the foreground in the binarizing step, and, for each label area to which the same label has been assigned in the first labeling step, A first label feature amount calculating step of calculating an image feature amount of the label region; and a first attribute determination of determining an attribute of the label region from the label feature amount for each label region calculated in the first label feature amount calculating step. Using the attribute determination result in the attribute determination step for each label area labeled in the first labeling step. An image processing method characterized by having a segment correction step for correcting the segment in the extracted. 26. A line segment removing step for removing a line segment extracted in the line segment extracting step from the image data binarized in the binarizing step, and the image data is set as a foreground in the binarizing step. A second labeling step of assigning the same label to the pixels and connected pixels from which the line segment has been removed in the line segment removing step; and a label region labeled in the first labeling step and the second labeling step. An inclusion relation determining step of determining an inclusion relation of each of the label areas labeled in the first step; an attribute determination result in the first attribute determining step for the label area labeled in the second labeling step; 2 from the inclusion determination results in the inclusion relationship determination step for each of the label areas labeled in the two labeling steps. Determining the attribute of the label area labeled in the first labeling step
An attribute determining step, wherein in the line segment correcting step, using the attribute determination result in the second attribute determining step for each label region labeled in the first labeling step, the line segment extracting step The image processing method according to claim 25, wherein correction of the extracted line segment is performed. 27. An edge extracting step of extracting an edge from the input image data; and an edge for removing the edge extracted in the edge extracting step from the image data binarized in the binarizing step. A removing step; a second labeling step of assigning the same label to the connected pixels which are foreground in the binarizing step and the edges of which are removed in the edge removing step; and the first labeling step An inclusion relation determining step of determining an inclusion relation between the label area labeled in step 2 and the label area labeled in the second labeling step; and the first attribute determining step for the label area labeled in the second labeling step. In the first and second labeling steps, A second attribute determining step of determining an attribute of the label area labeled by the first labeling step from the inclusion determination result in the inclusion relation determining step of each of the labeled label areas; In the minute correction step, the line segment extracted in the line segment extraction step is corrected using the attribute determination result in the second attribute determination step for each label area labeled in the first labeling step. 26. The image processing method according to claim 25, wherein: 28. An edge extracting step of extracting an edge from the input image data, and an image synthesizing step of synthesizing the line segment extracted in the line segment extracting step and the edge extracted in the edge extracting step. An edge / segment removing step of removing edges and line segments synthesized in the image synthesizing step from the image data binarized in the binarizing step; A second labeling step of assigning the same label to the connected pixels that have undergone the edge and line segment removal in the edge / line segment removal step, and a label area labeled in the first labeling step. An inclusion relation determining step of determining an inclusion relation of each of the label areas labeled in the second labeling step; An attribute determination result in the first attribute determination step for the label region labeled in the labeling step, and an inclusion determination result in the inclusion relationship determination step for each of the label regions labeled in the first and second labeling steps. From the second, the attribute of the label area labeled in the first labeling step is determined.
An attribute determining step, wherein in the line segment correcting step, using the attribute determination result in the second attribute determining step for each label region labeled in the first labeling step, the line segment extracting step The image processing method according to claim 25, wherein correction of the extracted line segment is performed. 29. The method according to claim 29, further comprising: a first small area removing step of removing a label area having a predetermined size or less from the label feature quantity for each label area calculated in the first label feature quantity calculating step. 29. The image processing method according to claim 25, wherein in the determining step, an attribute of the label area not removed in the first small area removing step is determined. 30. A second label feature amount calculating step for calculating an image feature amount for each of the regions labeled in the first labeling step; and a label area size calculated in the second label feature amount calculating step is a predetermined size. 31. The image processing method according to claim 25, further comprising: a second small area removing step of removing an area whose label area size is equal to or smaller than a predetermined size, as compared with. 31. A dilation processing step of dilating an extraction area obtained by synthesizing in the image synthesizing step by a predetermined size, wherein in the edge / line segment removing step, the dilation processing is performed in the dilation processing step. 31. The image processing method according to claim 28, wherein the extracted region is removed.