JP2584556B2 - How to set the reference density point - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for setting a highlight point and / or a shadow point as a reference density point in a gradation conversion apparatus used in the field of image processing such as a color scanner for plate making.

[0002]

2. Description of the Related Art A method for automatically setting highlight density and shadow density in a color separation apparatus such as a color scanner for plate making is disclosed in Japanese Patent Application Laid-Open No. 2-157775 by the present applicant.
No. 8 discloses a technique disclosed in Japanese Patent Publication No. The method disclosed in this conventional publication obtains a highlight density value and a shadow density value on the input side of a gradation conversion curve by a procedure as shown in FIG.

First, in step S501, an original image to be copied is prescanned, and a density value for each color component of the original image is obtained for each pixel.

Next, in step S502, the density values for each color component are averaged for each pixel to obtain an average density value.
Seek Nehi Sutoguramu.

Next, in step S503, a cumulative density value is calculated for each class for each color component, and a cumulative density value histogram as shown in FIG. 19 is obtained (however, FIG.
Is shown only for the cumulative density value histogram for

Next, in step S504, the relative frequency of the pixel obtained by cumulatively adding the pixels from the lower density class is obtained.
As shown in FIG. 18, a cumulative relative frequency histogram that changes from 0 to 100% in the range of the minimum and maximum generated density values is obtained.

Next, in step S505, predetermined cumulative density appearance rates RNH and RNH corresponding to highlight points and shadow points that provide the optimum gradation conversion characteristics empirically obtained.
By applying NS to the cumulative relative frequency histogram, a provisional highlight average density value DMH and a shadow average density value DMS corresponding to the cumulative relative frequency are obtained.

Finally, in step S506, the provisional highlight average density value DMH and the shadow average density value D
The MS is applied to the cumulative density value histogram for each color component shown in FIG. 19 to determine the input-side highlight density value and the input-side shadow density value for each color component. The gradation conversion curve should pass through the input-side highlight density value and the input-side shadow density value obtained as described above, and the output-side highlight density value and the output-side shadow density value arbitrarily set in advance. A highlight point and a shadow point, that is, a reference density point are set.

[0009]

However, according to the method disclosed in JP-A-2-157758, if the original image is a scene having a bright background or if the original image is a scene shot in backlight, In some cases, the input highlight density value was set to a value lower than the desired value, and the reproduced image had a dark finish. Conversely, if the original image is a scene with a dark background, the input-side shadow density value is set to a higher value, resulting in a whitish finish.

This is due to the following reasons. That is, for example, when the original image is a scene having a light background, the cumulative density value histogram shown in FIG. 19 is a histogram having a relatively low density and a high frequency due to the influence of the light background. Therefore, the provisional highlight average density value obtained from the cumulative relative frequency histogram of FIG. 18 becomes low, and the input-side highlight density value obtained in step S506 also becomes low. In this manner, when the input side highlight density value is lower, the tone conversion curve is set based on the input side the highlight side density value shifts to a highlight area output shadow density value side, reproductions as a result The result is a dark finish.

If the original image has a dark background, the cumulative density value histogram shown in FIG.
The input-side shadow density value obtained in 506 becomes higher. Accordingly, the gradation conversion curve is a transition of the shadow area to the output-side highlight density value side, and the reproduced image gradation-converted according to the gradation conversion curve has a whitish finish.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a method of setting a reference density point of a gradation conversion curve capable of obtaining a desired finish of a reproduced image without being affected by a background state of an original image.

[0013]

According to a first aspect of the present invention, there is provided a method of setting a reference density point, comprising the steps of: obtaining a gradation based on image data obtained by reading a duplication original image having a gradation; A method for setting a reference density point through which a gradation conversion curve to be set for performing conversion is to pass, wherein a first step of dividing an original image into a plurality of blocks, and A second step of obtaining a density histogram for each block indicating a relationship between a density value for each block and the number of pixels giving the density value, and a step of: To select a block group consisting of consecutive blocks
And a fourth step of detecting, from the block group selected in the third step, a block of a block group including a block having a density that is continuous with a predetermined block as a background block, and the number of pixels of the background block. Is relatively compressed for each class of the block-by-block density histogram with respect to the number of pixels of the block not selected as the background block, and after the fifth step, for each class, A sixth step of obtaining a density histogram of the entire original image by adding the number of pixels of all blocks, a seventh step of obtaining a cumulative density value histogram based on the density histogram of the entire original image, An eighth step of setting a reference density point based on the density value histogram. The cumulative density value histogram may include at least a highlight portion and a shadow portion in the entire original image.

According to a second aspect of the present invention, in the method of the first aspect, the predetermined blocks in the fourth step are blocks at four corners of the original image.

According to a third aspect of the present invention, in the method for setting a reference density point, the third step comprises the steps of:
When the difference between the average density values of blocks that are close to each other and have uniform densities is equal to or smaller than a predetermined value, the densities of these blocks are assumed to be continuous.

According to a fourth aspect of the present invention, in the method of setting a reference density point according to any one of the first to third aspects, the number of pixels of each class in the density histogram of the background block is less than one. And the number of pixels of each class of the density histogram of the blocks other than the background block by 1
, The number of pixels of the background block is compressed relative to the number of pixels of blocks other than the background block for each class of the density histogram.

According to a fifth aspect of the present invention, in the method of setting a reference density point of the fourth aspect, a coefficient for multiplying the number of pixels of each class in the density histogram of the background block is 0. .

According to a sixth aspect of the present invention, there is provided a method for setting a reference density point.
6. The method of setting a reference density point according to claim 1, wherein: (1) the density value is an average density value obtained by averaging the density values of the respective color components for each pixel; (2) the density value Is the density value of each pixel of the selected color component, or
(3) The density value is a density value of each color component for each pixel.

[0019]

According to the method of setting a reference density point according to the first aspect, a block which is uniformly bright or uniformly dark and has a density directly or indirectly continuous with a predetermined block. Are selected as background blocks. Also, the number of pixels of each background block is compressed relatively to the number of pixels of each block other than the background block, so that the density of the background block in the number of pixels of each class in the density histogram of the entire original image The proportion occupied by the number of pixels of each class in the histogram is reduced as compared with the case where there is no compression. Accordingly, when the background of the original image is a bright background, the input side highlight density value becomes larger compared to the case where the number of pixels of the background block is not relatively compressed. Move to the output side highlight density value side. Also, when the background of the original image is a dark background, the tone conversion curve is mainly represented by the input-side shadow density value being smaller than when the number of pixels of the background block is not relatively compressed. Move to the density value side.

According to the reference density point setting method of the second aspect, the density of the background block detected in the fourth step is directly or indirectly continuous with the four corner blocks of the original image.

In the method of setting a reference density point according to a third aspect of the present invention, the difference between the uniformly bright blocks or the uniformly dark blocks having uniform and close densities is equal to or less than a predetermined value. When, the block is a block having a continuous density.

According to the reference density setting method of the present invention, by reducing the number of pixels of a block detected as a background block, the number of pixels of this background block is reduced with respect to the number of pixels of blocks other than the background block. It is relatively compressed.

According to the reference density point setting method of the present invention, the pixels of the block selected as the background block are completely ignored in obtaining the cumulative density value histogram of the entire original image.

According to the method of setting the reference density point, the influence of the pixels in the background block is reduced according to the ratio of the color components near the reference density.

According to the method of setting a reference density point according to the present invention, when the background block is composed of pixels occupied by a specific color component, the influence of the pixels of the background block on the color component is reduced.

According to the method of setting a reference density point, the influence of the pixels of the background block is reduced for each color component.

[0027]

FIG. 10 is a schematic block diagram of a plate making scanner to which an embodiment of the present invention is applied. In the figure,
The image of the original image 100 is read for each pixel by the scanning reading device 200, and the image signal thus obtained is transferred to the image processing device 300. The image processing device 300
Highlight / shadow point setting unit 4 with functions described later
00 to perform processing such as setting a highlight / shadow point on an input image signal. Then, the processed image signal is provided to the scanning recording device 500. The scanning recording apparatus 500 converts an image signal into a dot signal, and, based on the converted signal, exposes and records a dot image on the photosensitive film 600.

FIG. 11 shows an image processing apparatus 300 including the highlight / shadow point setting section 400 described above. Here, first, the image data obtained by the scanning reading device 200 is read into the image memory 401 constituting the highlight / shadow point setting unit 400. The image memory 401 stores the original image OG in a large number of pixels, for example,
The image data is decomposed into 262144 pixels each having 512 rows and stored. C constituting the highlight / shadow point setting unit 400
The PU 402 converts the original image OG having a large number of pixels into N blocks T of vertical V and horizontal H as shown in FIG.
Divide into n. As described above, the original image having 262144 pixels is divided into N = 1024 blocks Tn where V = 32 and H = 32. In this case, one block Tn is composed of a total of 256 pixels, 16 pixels vertically and 16 pixels horizontally. Further, the CPU 402 calculates a highlight point and a shadow point of a gradation conversion curve based on the original image data read into the image memory 401 in a procedure described later.

The highlight point / shadow point calculated by the CPU 402 is given to a storage unit such as a look-up table provided in the gradation conversion device 301. Thereafter, the uncorrected digital density signal given from the scanning reading device 200 to the tone conversion device 301 is converted into a normalized digital density signal according to a tone conversion curve set based on the highlight point / shadow point. Is converted. This normalized digital density signal is subjected to a predetermined color operation in a color operation block 302 and then output to the scanning recording device 500. The color calculation block 302 performs a predetermined color calculation according to an instruction from an input / output unit (not shown) including a CRT, a keyboard, and the like.

The highlight / shadow point setting section 400
Sets a highlight point and a shadow point of the gradation conversion curve as follows. FIG. 1 shows a schematic setting procedure of the highlight point and the shadow point.

As shown in FIG. 1, the highlight / shadow point setting section 400 first detects a background portion BK of the original image OG as shown in FIG. 12 in step S1. In FIG. 12, the background portion BK is a portion other than the subject portion MA indicated by oblique lines.

Next, in step S2, the number of pixels of the background portion BK detected in step S1 is determined by subtracting the number of pixels of the background portion B
A cumulative relative frequency histogram of the entire original image OG is created in a state where the number of pixels in blocks other than K (the subject part MA in FIG. 12) is compressed relatively.

In step 3, a histogram of cumulative density values for each color component is created.

In step S4, based on the cumulative relative frequency histogram of the entire original image OG created in step S2 and the cumulative density value histogram created in step S3, a reference density point, that is, a highlight point, of the gradation conversion curve is obtained. And a shadow point are calculated.

The following steps S1 to S4
Will be described in detail.

In the step S1, FIGS.
The background portion BK in the original image OG is detected by the procedure shown in FIG.

First, in step S101 of FIG.
The original image OG is divided into N (H × V) blocks Tn of V length and H width (see FIG. 12). n is the block T
n is the block number and is an integer from 1 to N.

Next, steps S102 to S10
5, the density histogram hnj of each block Tn,
The density average value Dn and the density variance value Sn are sequentially obtained from n = 1 to N. Here, j is the class number of the histogram and 0
To J. When the maximum density value is Dmax and the density step is ΔD, J = Dmax
/ △ D. Hereinafter, the density histogram hnj, the density average value Dn, and the density variance value Sn will be described in more detail.

The density histogram hnj in this embodiment is a density histogram of an average density value obtained by averaging the density values of each color component of each pixel, and is created by the following procedure, for example. That is, the density values DR, DG, and DB for each color component obtained by pre-scanning the original image OG are averaged by the equation (1) to obtain an average density value DM for each pixel. Ask for.

[0040]

(Equation 1)

Then, for each block Tn, the average density value DM of the pixels belonging to the class having a predetermined width and the average density value D
The density histogram hnj is created as a histogram indicating the relationship with the number of pixels giving M.

The average density value Dn is the average value of the average density values DMni for the pixels included in each block Tn. That is, the density average value Dn of each block Tn can be obtained by the following equation (2). In the equation (2), I is the number of pixels in the block Tn.

[0043]

(Equation 2)

Next, in this embodiment, the density variance value Sn is determined by the following equation as a sample standard deviation using the average density value DMni of each pixel as a sample random variable.

[0045]

(Equation 3)

As described above, the density histogram hn
j, the density average value Dn, and the density variance value Sn are determined, and in steps S106 to S111 shown in FIG.
A block having a uniform density is detected from the blocks T1 to TN. In this embodiment, first, in step S107, the variance value Sn of the block Tn is compared with a predetermined value ε. Here, a block having a variance value Sn smaller than a predetermined value ε is a block having a uniform density. Blocks having a variance value Sn equal to or larger than a predetermined value ε are blocks having uneven density.

Then, for a block having a uniform density, the flag FSn is set to "1" in step S108.
For blocks with non-uniform densities, step S
At step 109, the flag FSn is set to "0". By repeating the routine of steps S107 to S111 until the block numbers n = 1 to N, the flags FSn of all the blocks Tn are set to "1" or "0".

Next, as shown in FIGS. 3 and 4, in steps S112 to S137, the continuity of the density of a block having a uniform density with the adjacent block is evaluated. The continuity of the density is evaluated using a scanning mask as shown in FIG. The scan mask shown in FIG. 13 includes a block Tn to be evaluated, a block Tn-H already evaluated above the evaluation block Tn, a block Tn-1 evaluated before the evaluation block Tn, and Block Tn-H- evaluated before block Tn-H
Covers 1

In step S112, FIG.
Prior to performing the evaluation using the scanning mask, the block numbers n of the blocks T1 to TN, the table numbers k of the later-described active tables ETk0 and ETk1, and the block T
All the labels Lc sequentially assigned to n are initialized to "1".

Next, in steps S113 to S114, it is determined whether or not the evaluation block Tn and the block Tn-H-1 covered by the scanning mask in FIG. 13 are blocks having uniform densities. If the blocks Tn and Tn-H-1 are both blocks having uniform densities, the difference between the average density Dn of the block Tn and the average density Dn-H-1 of the block Tn-H-1 is determined in step S115. Is compared with a predetermined value d. If the absolute value is smaller than the predetermined value d, the label Ln of the block Tn is determined in step S116.
To the same value as the label Ln-H-1 of the block Tn-H-1. When the label Ln is set to the same value as the label Ln-H-1, the process proceeds to step S136 described below.

In step S114, the block Tn-H-
If it is determined that the density in Step 1 is not uniform, Step S1
When it is determined in 15 that the absolute value of the difference between the average density value Dn and the average density value Dn-H-1 is equal to or greater than the predetermined value d,
Proceed to step S117.

In step S117, it is determined whether or not the block Tn-H is a block having a uniform density. If the block Tn-H is a block having a uniform density, in step S118, the absolute value of the difference between the average density Dn of the block Tn and the average density Dn-H of the block Tn-H is compared with a predetermined value d. . The density average value Dn and the density average value Dn−
If the absolute value of the difference from H is smaller than the predetermined value d, the label Ln of the block Tn is changed to the block Tn- at step S119.
Set the same value as the label Ln-H of H.

As described above, when the label Ln of the block Tn to be evaluated is set to the same value as the label Ln-H of the block Tn-H above it, in the next step S120, the immediately preceding block Tn of the evaluation block Tn is obtained. It is determined whether or not the block Tn-1 is a block having a uniform density. If this block Tn-1 is a block having a uniform density, step S12
In step 1, it is determined whether the value of the label Ln-H of the block Tn-H is equal to the value of the label Ln-1 of the block Tn-1. If not, the process proceeds to step S122.

In step S122, the block T
The absolute value of the difference between the average density Dn-H of nH and the average density Dn-1 of the block Tn-1 is compared with a predetermined value d. If the absolute value of the difference between the density average value Dn-H and the density average value Dn-1 is smaller than the predetermined value d, in step S123, the equitable tables ETk0 and ET for recording labels that can be integrated with each other.
The label Ln-1 and the label Ln-H are respectively recorded in k1.

That is, the blocks Tn and Tn-1
, Block Tn-H are all blocks of uniform density, label Ln-H and label Ln have the same value, and if block Tn-H and block Tn-1 have the same density, label Ln -1 can also be a label Ln-H and a label Ln.
0, recorded in ETk1. As described above, the labels Ln-1 and Ln-H are added to the equitable tables ETk0 and ETk1.
Is recorded, the table number of the EQUIV table is advanced by one in step S124, and then in step S136
Proceed to.

In step S120, the block Tn-1
Is not a block with a uniform density, step S121
If the label Ln-1 and the label Ln-H are already the same value at this stage, and if the density average value D
If the absolute value of the difference between nH and the average density value Dn-1 is equal to or greater than the predetermined value d, the process directly proceeds to step S136 without recording anything in the equitable tables ETk0 and ETk1.

When it is determined in step S117 that the block Tn-H is not a block having a uniform density, and in step S118, the absolute value of the difference between the average density value Dn and the average density value Dn-H is predetermined. If it is determined that the value is equal to or larger than the value d, the process proceeds to step S125.

In step S125, block T
It is determined whether or not n-1 is a block having a uniform density.
If the block Tn-1 is a block having a uniform density, in step S126, the absolute value of the difference between the density average Dn-1 of the block Tn-1 and the density average Dn of the evaluation block Tn is set to a predetermined value d. It is determined whether it is smaller than. If the absolute value is smaller than the predetermined value d, the label Ln of the block Tn is changed to the block Tn- at step S127.
The same value as the label Ln-1 of 1 is set. After setting the label Ln to the same value as the label Ln-1, the process proceeds to step S136.

In step S125, block Tn-1
Is determined not to be a block having a uniform density, or when it is determined in step S126 that the absolute value of the difference between the average density value Dn-1 and the average density value Dn is equal to or greater than a predetermined value d,
That is, although the evaluation block Tn is a block having a uniform density, the continuity of density of any of the other blocks Tn-H-1, Tn-H and TN-1 covered by the above-described scanning mask of FIG. If it is determined that there is no
Proceed to 128.

In step S128, a new label Lc that has not been used is assigned as the label Ln to be attached to the evaluation block Tn. As described above, when a new label Lc is used, the value of the label Lc is advanced by one in step S129. After updating the value of the label Lc in this way, the process proceeds to step S136.

In step S113, the evaluation block Tn
If it is determined that is not a block having a uniform density, the process proceeds to step S130. In steps S130 to S133, when the evaluation block Tn is not a block having a uniform density, the label Ln-1 of the block Tn-1 before the evaluation block Tn and the label of the block Tn-H above the evaluation block Tn It is determined whether Ln-H is a block that can be integrated.

That is, in steps S130 and S131, both the blocks Tn-H and Tn-1 are blocks having uniform densities, and in step S132 the values of the labels Ln-H and Ln-1 are not the same. And step S13
In step 3, the density average value Dn-1 of the block Tn-1 and the block Tn-
When the absolute value of the difference from the density average value Dn-H of H is smaller than the predetermined value d, the label Ln-
1 and the label Ln-H are recorded in the equitable tables ETk0 and ETk1 as labels that can be integrated with each other. Then, in step S135, the numbers of the equitable tables ETk0 and ETk1 are updated by one, and the flow advances to step S136.

In steps S130 and S131, block T
If one of the block n-1 and the block Tn-H is not a block having a uniform density, the label L is determined in step S132.
If the values of n-1 and Ln-H are already the same, or in step S133, the density average value Dn-1 and the density average value Dn
If the absolute value of the difference from nH is equal to or greater than the predetermined value d, the process proceeds to step S136 at that time.

In step S136, the block number of the evaluation block Tn covered by the scanning mask of FIG. 13 is advanced by one. If the updated block number n is equal to or less than N in step S137, the process returns to step S113 to return to step S113. The routine from S113 to S136 is repeated. By the above routine, labels are applied to all blocks having a uniform density. In step S137, the number n of the evaluation block Tn
If N exceeds N, the process proceeds to step S138.

In steps S138 to S144 shown in FIG.
The label recorded in Tk1 is integrated.

First, in step S138, the table number k of the active tables ETk0, ETk1 is returned to the previous number, and in step S139, the previous active table ETk0, ETk1 is returned.
It is determined whether the table number k of k0 and ETk1 is 1 or more. If the table number k is less than 1 in step S139, that is, if the table number k is “0”, the equiv table
Since ETk1 has not been used, the process proceeds to step S145 in FIG.

If the number k is equal to or greater than 1 in step S139, the label Ln of the block Tn is compared with one of the active tables ETk0 in steps S140 to S144 (step S141), and the label stored in the active table ETk0 is compared. Is equal to the label Ln, the label of the block recorded in the other equitable table ETk1 is rewritten to the label Ln (step S142).

When the processing of steps S141 and S142 is completed for the blocks T1 to TN, step S
At 143, one block number n is updated. The routine of steps S141 to S143 is performed in step S144.
Is repeated until n = N. If n> N is satisfied in step S144, the process returns to step S138 to return the table number k of the even tables ETk0 and ETk1 to the immediately preceding number. In step S139, the table number k becomes 1
Equivalent table ETk0, E until less than 0
The routine of steps S140 to S144 is repeated for Tk1.

Next, in this embodiment, in the routine after step S145 shown in FIG. 6, the same labels as those of the blocks T1, TH, TN-H + 1, TN (see FIG. 12) at the four corners of the original image are given. The block is evaluated as a block constituting the background. That is, in this embodiment, it is assumed that the four corners of the original image are background portions. In steps S145 to S148, all flags BGn indicating whether the block Tn is a background block are set to the block Tn.
Is a routine for clearing to "0". Hereinafter, the routine after step S149 will be described.

First, in step S149, the search number m of the background block is set to 1. Step S150-S
In 153, the blocks at the four corners T1, TH, TN-H-1,
It is sequentially determined whether or not TN is reached. When the search number m = 1, in step S150, the block T1 is searched as one of the background evaluation reference blocks T1, TH, TN-H-1, and TN, and the process proceeds to step S154.

In the next steps S154 to S159, the label Lm, that is, the label L1 of the background evaluation reference block T1 searched in step S150 is replaced with all the blocks Tn.
Are sequentially compared with the label Ln (step S155).
If Ln = L1, the flag BGn of the block Tn is set to "1" in step S156. If Ln ≠ L1, the flag BGn of the block Tn is set to “0” in step S157. Step S156 or step S15
When the flag BGn is set to "1" or "0" in step 7, the process proceeds to step S158 to update the block number n by one. The routine of steps S155 to S158 is repeated until the block number n = N in step S159. If n> N in step S159, step S1
Proceed to 60.

In step S160, one search number m is updated, and m = 2. In step S161, the search number m
= 2 is compared with the number N of blocks. Since 2 ≦ N, the process returns to step S150.

Thereafter, the routine of steps S150 to S161 is repeated for search numbers m = 2 to N. However, when the search number m is m ≠ 1, H, N−H + 1, N, the process directly proceeds from step S153 to step S160, and one search number m is updated.

Search number m = H, N−H + 1, or N
In this case, the process proceeds to step S154 as in the case of m = 1, and the routine of steps S155 to S159 is repeated. That is, when m = H, the flag BGn of the block Tn of Ln = LH is set to “1”, while the flag BGn of the block Tn of LnnLH is set to “0”. Also, m
= N−H−1, the flag BGn of the block Tn of Ln = LN−H−1 is set to “1”, while the flag BGn of the block Tn of Ln ≠ LN−H−1 is set to “0”. Furthermore, m
= N, the flag BGn of the block Tn of Ln = LN
Is set to “1”, while the flag BGn of the block Tn of Ln ≠ LN is set to “0”.

As described above, all blocks T1 to T
The N flags BG1 to BGN are set to "1" or "0". However, the background evaluation blocks T1, TH, TN-H-1, T
The flag BGn of the block Tn to which the same label as one of the blocks of N is attached is set to "1". That is, when the flag BGn is set to "1" or "0" for one block Tn in the routine of steps S149 to S161, the flag BGn of the block Tn is set to "1". Prioritize.

As described above, steps S101 to S1
A routine T61 detects a background block Tn from among the blocks T1 to TN. After detecting the block as the background in this way, a cumulative relative frequency histogram of the entire original image is created in steps S201 to S223 shown in FIGS.

First, in steps S201 to S213, the density histogram hnj of each of the blocks T1 to TN is set.
The density histogram of the entire original image is created based on each class j. In steps S201 to S204, the number of pixels DFj corresponding to the class number j (= 0 to J) of the density histogram hnj is cleared to zero.

Next, in step S205, the block number n is set to "1", the class number j is set to "0", and the total number of pixels y is set to "0". Thereafter, step S206-
By repeating the routine of S213 until n> N in step S213, the number of pixels DFj and the total number of pixels y for each class in the entire original image are obtained. Step S20
The routine from 6 to S213 is as follows.

First, in step S206, it is determined whether or not the flag BGn of the block Tn is "1". If the flag BGn is "1", that is, if the block Tn is a block evaluated as a background, a coefficient α = α0 is selected in step S207. Coefficient α
0 is 0 ≦ α0 <1. If the flag BGn is "0", that is, if the block Tn is a block evaluated as not being a background, a coefficient α = 1 is selected in step S208.

Next, in steps S209 to S211, (1) the number of pixels DFj of each class of the image composed of the blocks T1 to Tn and (2) the blocks T1 to T
The total number of pixels y of the image consisting of n is obtained.

The above-described routine of steps S206 to S211 is repeated until n> N is satisfied in step S213 as a result of updating one block number n in step S212.

Next, the number of pixels DFj of each class and the total number of pixels y obtained in steps S209 to S211 will be described in detail.

When the block number n = 1 and the class number j = 0, in step S209, the corrected pixel number α1 of the corresponding class is added to the pixel number DF0 cleared in step S202.
DF0 + α1 obtained by adding h10. H10 is the number of pixels D corresponding to the class number "0" in the image of the block T1.
It is calculated as F0. Here, α1 is set in step S206.
係数 is the coefficient α of the block T1 selected in S208.
In this case, since the number of pixels DF0 is “0” in step S202, the number of pixels DF0 calculated in step S209 is
Is α1 · h10.

After calculating the number of pixels DF0, step S2
At 10, one class number j is updated, and the updated class number j is compared with the final class number J at step S211. If j ≦ J in step S211, the process returns to step S209, and the routine of steps S209 to S211 is repeated. Thus, the number of pixels D of all classes when n = 1
F0 to DFJ are sequentially obtained.

Further, the block number n = 1 and the class number j =
In step S209, in step S209, the total pixel number y cleared in step S205 is added to the corrected pixel number α of the class number j = 0.
Calculate y + α1 · h10 by adding 1 · h10. Since the total pixel number y is “0” in step S205, the total pixel number y calculated in step S209 when the class number j = 0 is α1 · h10.

Next, in step S210, one class number is updated to j = 1. In step S211, 1 ≦
J returns to step S209. Next step S20
9, the total number of pixels y calculated when the class number j = 0 is y =
α1 · h10, corrected pixel number α1 when class number j = 1
H11 is added, and the total number of pixels y when the class number j = 1 is y =
Calculate α1 (h10 + h11). Hereinafter, similarly, j =
By repeating the steps S209 to S211 in steps 2 to J, the total number of pixels y in the block T1 is calculated as y =
α1 (h10 + h11 +... h1J) is calculated.

If j> J in step S211, the flow advances to step S212 to update the block number n by one and set n = 2. Then, since 2 ≦ N in step S213, the process returns to step S206. Next steps S206 to S20
In step 8, n = 2, that is, the coefficient α2 of the block T2 is selected.

In the next steps S209 to S211, n
By adding the corrected number of pixels α2 · h2j of the corresponding class of the block T2 to the number of pixels DFj of each class calculated at = 1 = α1 · h1j, it corresponds to the class number j in the image composed of the block T1 and the block T2. Α1 · h1j + α2 · h2j is determined as the number of pixels DFj. In step S209, the total number of pixels y = α1 (h
10 + h11 +... H1J) and the correction pixel number α2 · h2j is sequentially added. By repeating the routine of steps S209 to S211 until j> J in step S211, the total number of pixels y = α1 (h10 + h11 +... H1J) + α2 (h
20 + h21 +... H2J).

Thereafter, similarly, the routine of steps S206 to S213 is repeated with n = 3 to N, whereby the number of pixels DFj and the total number of pixels of each class of the entire original image as shown in equations (4) and (5) are obtained. The number of pixels y is calculated.

[0090]

(Equation 4)

[0091]

(Equation 5)

As described above, the density histogram of the entire original image can be created by obtaining the number of pixels DFj of each class in the entire original image. As is clear from the equations (4) and (5), the number of pixels of the background block detected in steps S101 to S161 is determined in step S2.
Since the coefficient α0 selected at 07 satisfies 0 ≦ α0 <1, the number of pixels in blocks other than the background is neglected or ignored.

Next, in steps S214 to S219, the cumulative number of pixels CDFj (C
DF1 to CDFJ). First, step S214
Then, the class number j is initialized to 0, and the process proceeds to step S215. If j ≠ 0 in step S215, that is, if j = 0, the process proceeds to step S216, and if j ≠ 0 in step S215, the process proceeds to step S217.

In step S216, the number of pixels DF0 of the class having the class number j = 0 is converted into the cumulative number of pixels CDFO. In step S217, the number of pixels DFj of class number j
Is added to the cumulative pixel number CDFj-1 up to the previous class j-1 to obtain the cumulative pixel number CDF up to the class number j.
Seeking j. In step S218, the class number j is updated by one. The updated class number j is the step S
If j ≦ J in 219, the process returns to step S215. If j> J in step S219 as a result of increasing the class number j by 1 in step S218, the next step S
Proceed to 220.

Steps S214 to S219 as described above
According to the routine, when the class number j = 1, the cumulative number of pixels CDF1 becomes the cumulative number of pixels CDF0 corresponding to the class number j = 1-1 = 0 and the pixel number DF corresponding to the class number j = 1.
The accumulated pixel numbers CDF2, CDF3,... C are sequentially obtained from the lower class number j.
DFJ is required.

In steps S220 to S223,
Each cumulative pixel count CDFj with respect to the total pixel count y of the entire original image
Is obtained. Step S22
At 0, the class number j is initialized to "0". In step S221, the relative frequency RNj (%) is obtained from the expression RNj = CDFj × 100 / y based on the cumulative pixel count CDFj obtained in steps S214 to S219 and the total pixel count y of the entire original image. A step S222 updates the class number j by one. Step S223 is step S2
The class number j updated at 22 is compared with the final class number J. Steps S221 to S until j> J
By repeating the routine of 223, the relative frequencies RNj of all the classes are obtained.

The cumulative relative frequency histogram as shown in FIG. 14 is obtained by setting the average density class value DMj (n = 0 to J) on the horizontal axis and the relative frequency RNj calculated in steps S220 to S223 on the vertical axis. Can be created. The cumulative relative frequency histogram shown in FIG. 14 is between 0% and 1 in the range between the minimum density DMmin and the maximum density DMmax.
The shape changes up to 00%. The cumulative relative frequency histogram of FIG. 14 is approximated by a curve by making the class width sufficiently small.

Next, steps S301 to S3 shown in FIG.
At 15, a cumulative density value histogram for each color component is created.

First, in steps S301 to S304, each block T obtained in step S103 in FIG.
n based on the density histogram hnj for each block T
n, cumulative density values DRnj, DGnj, and DBn for each color component
j. That is, for each class of the density histogram hnj obtained in step S103, the density value of each pixel included therein is extracted, and the density value of each color component is cumulatively added. This process is performed independently for each class. Next, in steps S305 to S313, the cumulative density value of the class number j in the entire original image is calculated.
In step S305, the class number j is initialized to 0. In step S306, DRj, DGj, and DBj are cleared as the cumulative density values for each color component of the entire original image. Next, in step S307, the block number n is set to 1.

The following steps S308 to S310 are the same as steps S206 to S208. That is, in step S308, it is determined whether or not the flag BGn of the block Tn is "1". If the flag BGn is "1", in step S309, α = α as a coefficient
Select 0. If the flag BGn is “0”, α = 1 is selected as a coefficient in step S310.

In step S311, the previously calculated cumulative density values DRj, DGj, and DBj are respectively set to α · DRn.
j, α · DGnj and α · DBnj. In step S312, the block number n is advanced by one. The block number n increased by 1 in step S313 is n
It is determined whether or not ≦ N. If n ≦ N, the process returns to step S308, and thereafter, n> n in step S313.
The routine of steps S308 to S313 is repeated until N is reached.

Thus, the cumulative density value DRj = α1 · DR1j + α2 · DR2j +... + ΑN for each class of the entire original image
· DRNj, DGj = α1 · DG1j + α2 · DG2j + ...
+ ΑN · DGNj, DBj = α1 · DB1j + α2 · DB2j +
... + ΑN · DBNj is calculated. Coefficient α1, α2 ……
αN is 1 or α0.

In steps S306 to S313, in step S314, the class value j is advanced by one, and in step S315, the class value j advanced by 1 is j ≦ J
Is determined, the process is repeated until the class value j> J in step S315. That is, in steps S305 to S315, the cumulative density values of the entire original image for all the classes from j = 0 to J are calculated, whereby the cumulative density value histogram for each color component of the entire original image as shown in FIG. Can be created.

The number of pixels Pj corresponding to each class in the cumulative density value histogram for each color component in FIG. 15 is obtained by using the number of pixels Pj corresponding to the class value DMj for each block Tn as follows. P1j + α2 P2j + ... + αN
-It can be obtained as PNj.

Next, a highlight point and a shadow point of the gradation conversion curve are obtained based on the cumulative relative frequency histogram of FIG. 14 and the cumulative density value histogram of each color component of FIG.

First, with respect to the cumulative relative frequency histogram of FIG. 14 created by the routine of steps S201 to S223, an optimum gradation conversion characteristic, for example, which is empirically obtained from a large number of sample original images prepared in advance, is obtained. By adopting the cumulative density appearance rates RNH and RNS corresponding to the given highlight point HL and shadow point SD, the provisional highlight average density value DMH and shadow average density value DMS corresponding to the highlight point HL and shadow point SD are obtained. Respectively. The values of the cumulative density appearance rates RNH and RNS are about 1% and 98%, respectively. The provisional highlight average density value DMH and shadow average density value DMS thus obtained
Is applied to the cumulative density value histogram for each color component shown in FIG.

In the cumulative density value histogram of FIG. 15, on the highlight side, the area (DMmin ≦ DM ≦ DMH) is lower than the temporary highlight average density DMH, and on the shadow side, the area is higher than the temporary shadow average density value DMS. Area (DMS ≦ D
The classes included in M ≦ DMmax are shown with diagonal lines. In the illustrated example, the provisional highlight average density value DM
H, average density value DMS, class value DM respectively
5, DM (J-2) is set.

For example, for the color component R, the cumulative density value DR1 corresponding to the class values DM1 to DM5 in the entire original image.
Add ~ DR5. Further, the numbers of pixels PR1 to PR5 in the range corresponding to the class values DM1 to DM5 are added. From these two added values, the input-side highlight density value D RH of the color component R is calculated according to Equation 6.
It is calculated by the following equation.

[0109]

(Equation 6)

The same processing is performed on the shadow side, and the input side shadow density value DRS is obtained by the equation (7).

[0111]

(Equation 7)

The cumulative density values DR (J-2), DR (J-1), DRJ and the number of pixels PR (J-2), PR (J-1), PRJ are classified into classes as in the case of the equation (6). It is provided corresponding to the values DM (J-2), DM (J-1), and DMJ.

Although not described in detail here, the input-side highlight density values DGH and DBH and the input-side shadow density value DG are similarly applied to the other color components G and B.
S and DBS can be obtained.

Next, the setup executed using the input-side highlight density values DRH, DGH, and DBH and the input-side shadow density values DRS, DGS, and DBS obtained as described above will be described.

Now, it is assumed that the original image is a scene having a bright background. In this case, the cumulative relative frequency histogram HI indicated by a solid line in FIG. 14 created by the routine of steps S201 to S223 in the above-described embodiment is obtained by creating a background portion under the same conditions as other portions by a conventional method.
As a whole, compared to the cumulative relative frequency histogram HI ′ indicated by the one-dot chain line in FIG. This is because, in the embodiment according to the present invention, when calculating the cumulative number of pixels CDFj of each class in the entire original image in step S209, steps S101 to S101 are performed.
This is because the coefficient α0 smaller than 1 is multiplied by the number of pixels of each class of the block Tn of the background portion detected at 161. That is, by multiplying the number of pixels of each class of the block Tn of the bright background by a coefficient α0 smaller than 1, the level of the class with low density in the average density histogram of the entire original image is suppressed, and That is, the cumulative relative frequency histogram is biased toward the high density side as a whole.

The cumulative relative frequency histogram RNH on the highlight side is added to the cumulative relative frequency histogram shown in FIG.
Is applied, the provisional highlight average density value DMH obtained when applied to the cumulative relative frequency histogram HI is
It is larger than the provisional highlight average density value DMH 'when applied to the cumulative relative frequency histogram HI'. Therefore, based on the tentative highlight average density value DMH, the input-side highlight density value DRH of the color component R given by the equation (6) and the input-side highlight density of the color components G and B given by the same equation The values DGH and DBH are larger than the input-side highlight density values DRH ', DGH' and DBH 'for each color component obtained based on the provisional highlight average density value DMH' according to the conventional method.

As shown in FIG. 14, the cumulative relative frequency histogram HI indicated by the solid line created by the method of the above embodiment.
And the cumulative relative frequency histogram HI ′ indicated by the dashed line formed by the conventional method almost coincides before the relative frequency is 100%. Therefore, the provisional shadow average density value DMS obtained when the cumulative density appearance rate RNS on the shadow side is applied to the cumulative relative frequency histogram created in the above embodiment, and the cumulative density appearance value DMS obtained in the cumulative relative frequency histogram created by the conventional method. This is very close to the tentative shadow average density value DMS 'obtained when the rate RNS is applied. Therefore, hereinafter, for convenience, the provisional average shadow density values DMS and DMS 'are assumed to be equal. Further, the shadow-side density value DRS of the color component R obtained by the equation (7) based on the temporary shadow average density value DMS
And the shadow-side density values DGS and DBS of the color components G and B are
It is assumed that they are respectively equal to the shadow-side density values DRS ', DGS', and DB 'obtained based on the temporary shadow average density value DMS'.

FIG. 16 shows a gradation conversion curve GCR set using the input-side highlight density value DRH and the shadow-side density value DRS of the color component R obtained by the method of the above embodiment, and the highlight obtained by the conventional method. 7 is a graph showing a tone conversion curve GCR 'set using the side density value DRH' and the shadow side density value DRS '. The horizontal axis is the input density value DI input for each color component, and the vertical axis is the output density value DO. The output side highlight density value DOHL and the output side shadow density value DOSD are values that are fixed in common for all color components. Here, for convenience, all the input-side highlight density values D
RH, DGH, and DBH are the input-side highlight density values DRH ', DG
H 'and DBH', and the ratio of DRH =
It is assumed that DGH = DBH and DRH '= DGH' = DBH '. Also, the input side shadow density values DRS, DGS, and DBS are DRS = D
GS = DBS, and the input side shadow density values DRS ', DG
S 'and DB' are assumed to be DRS '= DGS' = DBS '.
In such a case, the gradation conversion curve GCG of the color component G and the gradation conversion curve GCB of the color component B become equal to the gradation conversion curve GCR of the color component R shown in FIG.
And GCB 'are equal to the gradation conversion curve GCR' shown in FIG.

As is apparent from FIG. 16, the highlight point HLR corresponding to the highlight-side density value DRH is set at a position shifted in parallel to the right from the highlight point HLR 'corresponding to the highlight-side density value DRH'. Is done. On the other hand, as described above, the shadow side density value DRS is equal to the shadow side density value DR.
Equal to S '. Accordingly, the shadow point SDR corresponding to the shadow side density value DRS coincides with the shadow point SDR 'corresponding to the shadow side density value DRS'.

In this case, as shown in FIG. 16, the gradation conversion curve GCR passes above the gradation conversion curve GCR ', that is, on the highlight side output set value DOHL side. As described above, here, the gradation conversion curve GCR = GCG = GCB
And the gradation conversion curve GCR '= GCG' = GCB '
Therefore, the gradation conversion curve GCG is equal to the gradation conversion curve GCG.
And the gradation conversion curve GCB passes through the highlight side output setting value DOHL from the gradation conversion curve GCB '. Therefore, even if the same input density value DI is given, the gradation conversion curves GCR, GCG,
The output density value DO converted by the GCB is higher than the output density value DO converted by the tone conversion curves GCR ', GCG', and GCB '.
Approach HL.

As is clear from the above description, when the original image is a scene having a bright background, the highlight-side density values DRH, DGH, and DBH and the shadow-side density values DRS, DGS, and DBS obtained in the above embodiment are obtained. When the setup is performed by using, the density value DR on the highlight side obtained by the conventional method is obtained.
H ', DGH', DBH ', shadow side density values DRS', DG
Compared to the case where setup is performed using S 'and DB',
As for the output density value DO, the highlight area approaches the highlight-side output set value DOHL, and a duplicated image with a brighter finish can be obtained.

On the other hand, although not described in detail, according to the above embodiment, when the original image is a scene having a dark background, the shadow side density values DRS, DGS, and DBS are changed to the shadow side density values DRS ', DGS 'and DBS' are set lower. Therefore, in this case, the gradation conversion curve GCR passes through the lower side of the gradation conversion curve GCR ', that is, the shadow side output set value DOSD, and is converted by the gradation conversion curve GCR even if the same input density value DI is given. Output density value DO
Is closer to the shadow-side output set value DOSD than the output density value DO converted by the gradation conversion curve GCR '. That is, by setting the shadow-side density values DRS, DGS, and DBS according to the above-described embodiment, even if the original image is a scene having a dark background, a duplicated image that is not as whitish as the conventional image can be obtained.

The gradation conversion curve does not necessarily have to be a straight line like the gradation conversion curve of the embodiment shown in FIG. That is, the gradation conversion curve may be an appropriate curve.

In the above embodiment, the density histogram for each block Tn obtained in step S103 is a density histogram of an average density obtained by averaging the density values of the respective color components for each pixel. Instead of this average density value,
A lightness value obtained by a weighted average of each color component density value may be used. Further, the density histogram may be a density histogram of the density for each color component. In this case, the color components R,
A density histogram may be obtained for all of G and B, or a density histogram may be obtained for only one color component selected in advance. If a density histogram is to be obtained for all of the color components R, G, B, step S20
1 to S223, the cumulative relative frequency histogram is also obtained for each color component, and the cumulative density appearance rate RNH is obtained for each color component.
, RNS.

Then, based on the cumulative density appearance rates RNH, RNS of the respective color components and the cumulative density value histogram obtained in steps S301 to S315, the highlight-side density values DRH, DGH, DBH and the shadow-side density values DRS, DGS ,
Ask for DBS. On the other hand, when the density histogram is obtained only for one color component selected in advance in step S103, the cumulative density value histogram obtained in steps S301 to S315 may be only the cumulative density value histogram of the corresponding color component.

In the above embodiment, step S11
5, in steps S118, S122, S126, and S133, the difference between the average density values of the two blocks having uniform densities and being close to each other is obtained, and when the difference is smaller than a predetermined value d, the density between the two blocks is calculated. Are determined to be continuous. However, steps S115, S1
In steps S18, S122, S126, and S133, the variance value of the density of the entire blocks is obtained, and when the variance value is equal to or less than a predetermined value, it is determined that the density between the two blocks is continuous. Is also good.

In the above embodiment, step S206
In steps S208 to S208, the coefficient α = α0 (α0 <1) when the block Tn is a block of the background portion, and the coefficient α = 1 when the block Tn is a block other than the background portion. When the block is a partial block, the coefficient α = 1. When the block Tn is a block other than the background part, the coefficient α = α0 ′ (α0 ′>
It may be 1). Even in this case, α0 = 1 / α0
', The relative frequency RNj obtained in step S221
Is the same as in the above embodiment.

Further, the present invention can be used not only in plate making scanners but also in copying machines and facsimile machines having gradation reproducibility.

The predetermined block is not necessarily a block at the four corners of the original image, but may be a block at the upper two corners, or may be a block other than a block at the corner in some cases.

Further, when the image information of the original image is stored in the large-capacity memory in advance, the pre-scan by the scanning reading device is unnecessary, and the image data may be read directly from the memory and used.

[0131]

According to the first aspect, when the background of the original image is a light background, the gradation conversion curve can be shifted to the output highlight density side as a whole, so that the duplicated image has a dark finish. Can be prevented. Also,
When the background of the original image is a dark background, the gradation conversion curve can be shifted as a whole to the output side shadow density side, so that the finish of the duplicated image does not become whitish.

According to the second aspect, when the original image is such that the object is drawn in the center portion and the periphery is the background portion, it is possible to obtain a duplicate image of an appropriate finish.

According to the third aspect, it is possible to determine whether or not the density between adjacent blocks is continuous only by calculating the average value of the density of each block.

According to the fourth aspect, the number of pixels of the background block can be relatively compressed based on the number of pixels of the blocks other than the background block.

According to the fifth aspect, the reference density point can be set with the influence of the background block completely removed.

According to the present invention, the reference density points of all the color components can be created only by obtaining one cumulative density value histogram of the entire original image.

According to the seventh aspect, it is only necessary to obtain the cumulative density value histogram of one color component, and the process of obtaining the reference density point can be simplified.

According to the eighth aspect, since the cumulative density value histogram is obtained independently for each color component,
An appropriate reference density point can be obtained for each color component.

[Brief description of the drawings]

FIG. 1 is a flowchart schematically showing a method of setting a reference density point according to the present invention.

FIG. 2 is a flowchart illustrating a method of detecting a background portion.

FIG. 3 is a flowchart illustrating a method of detecting a background portion.

FIG. 4 is a flowchart illustrating a method of detecting a background portion.

FIG. 5 is a flowchart illustrating a method of detecting a background portion.

FIG. 6 is a flowchart illustrating a method of detecting a background portion.

FIG. 7 is a flowchart illustrating a process of obtaining a cumulative relative frequency histogram.

8 is a flowchart showing a process of determining the cumulative relative frequency histogram.

FIG. 9 is a flowchart illustrating a process of obtaining a cumulative density value histogram for each color component .

FIG. 10 is a schematic block diagram of a plate making scanner to which an embodiment of the present invention is applied.

FIG. 11 is a block diagram illustrating an image processing apparatus.

FIG. 12 is an explanatory diagram showing a state where an original image is divided into a plurality of blocks.

FIG. 13 is an explanatory diagram showing a scanning mask.

FIG. 14 is a diagram showing a cumulative relative frequency histogram.

15 is a diagram showing the cumulative concentration Nehi Sutoguramu.

FIG. 16 is a graph showing a gradation conversion curve.

FIG. 17 is a flowchart illustrating a conventional method of setting a reference density point.

FIG. 18 is a diagram showing a cumulative relative frequency histogram obtained by a conventional method.

FIG. 19 is a diagram showing a cumulative density value histogram for each color component obtained by a conventional method.

[Explanation of symbols]

 400 highlight / shadow point setting unit Tn block hnj density histogram Dn density average value Sn density variance value DFj pixel count CDFj cumulative pixel count HI cumulative relative frequency histogram RNH, RNS cumulative density appearance rate DRH, DGH, DBH Input highlight density Value DRS, DGS, DBS Input side shadow density value HL Highlight point SD Shadow point GCR Tone conversion curve

Claims (8)

(57) [Claims] 1. A method of setting a reference density point through which a gradation conversion curve to be set for performing gradation conversion at the time of duplication based on image data obtained by reading a duplication target original image having gradations. A first step of dividing the original image into a plurality of blocks; and a density for each block indicating a relationship between a density value and the number of pixels giving the density value for each block based on image data of the original image. A second step of obtaining a histogram, a third step of selecting a block group of blocks which are close to each other, have uniform densities, and have continuous densities based on the image data, and a third step. A fourth step of detecting, as a background block, a block of a block group including a block whose density is continuous with a predetermined block from the block group selected in the step; Is relatively compressed for each class of the block-by-block density histogram with respect to the number of pixels of a block not selected as the background block. After the fifth step, for each class, A sixth step of obtaining a density histogram of the entire original image by adding the number of pixels of all blocks; a seventh step of obtaining a cumulative density value histogram based on the density histogram of the entire original image; An eighth step of setting a reference density point based on the density value histogram. 2. The method of setting a reference density point according to claim 1, wherein the predetermined block in the fourth step is a block at four corners of a rectangular original image. 3. The method of setting a reference density point according to claim 1, wherein the third step is such that a difference between average density values of blocks which are close to each other and whose density is uniform is equal to or less than a predetermined value. A method of setting a reference density point, characterized in that the density of these blocks is continuous at one time. 4. The method of setting a reference density point according to claim 1, wherein in the fifth step, the number of pixels of each class of the density histogram of the background block is multiplied by a coefficient less than one. At the same time, by multiplying the number of pixels of each class of the density histogram of the blocks other than the background block by 1, the number of pixels of the background block is relatively compressed with respect to the number of pixels of the blocks other than the background block for each class. Setting a reference density point. 5. The method of setting a reference density point according to claim 4, wherein a coefficient by which the number of pixels of each class of the density histogram of the background block is multiplied by zero. 6. The method of setting a reference density point according to claim 1, wherein the density value is an average density value obtained by averaging the density values of the respective color components for each pixel. How to set the reference density point. 7. The reference density setting method according to claim 1, wherein the density value is a density value for each pixel of the selected color component. How to set points. 8. The method of setting a reference density point according to claim 1, wherein the density value is a density value of each color component for each pixel. .