JPH0229880A - Separation of video rate area - Google Patents

Abstract

PURPOSE:To economically speed up labeling processing by reading out a table memory expressing the connecting relation of temporary label values allocated to respective picture elements at the initial cycle and writing data in the memory at the succeeding cycle so that temporary labeling can be executed by one access at maximum in each cycle inputting each picture element. CONSTITUTION:A video rate area separating circuit is provided with a temporary labeling circuit 1, a frame memory 4 for storing one frame of temporary labeled images, a connecting relation integrating circuit 3 for integrating the connecting relation of temporary label values on the basis of a connecting relation table 2, and a density converting circuit 5 for converting the density of respective picture elements in accordance with the temporary label values stored in the memory 4. The area of an image is separated, and at the time of a window pattern requiring reading out/writing data from/in the table 2 in one cycle inputting one picture element in the circuit 1, reading operation is executed in the initial cycle and writing operation is executed in the succeeding cycle. Consequently, temporary labeling can be executed by accessing the table 2 once at maximum in each cycle inputting each picture element.

Description

[Detailed Description of the Invention] [Table of Contents] Overview Industrial Application Fields Prior Art and Problems to be Solved by the Invention Means for Solving the Problems Actions Examples Effects of the Invention [Summary] Input at video rate Regarding the video rate area separation method, which attaches a unique label to each area of each image in order to clearly identify multiple images that exist independently within each screen, it is a high-speed image processing method that performs continuous image processing at the video rate. In the labeling processor, the connection relationship table memory is read once every time one pixel is input.

Two writes are required, and the connection relationship table memory is
With the aim of solving the problem of requiring ultra-high-speed memory elements, we created a screen with multiple images that exist independently within one screen. Window patterns (1) to (4), (6) to Q[I
), a frame memory that stores one frame of the provisionally labeled image, and a connection relationship between the provisional label values after the provisional labeling process for one frame is completed. a connection relationship integration table that integrates based on the above connection relationship table,
Based on the result of integrating the connection relationships, image regions are separated by a video rate region separation circuit equipped with a density conversion circuit that converts the density of each pixel in accordance with the provisional label value stored in the frame memory. At the same time, in the case of a window pattern (αQ to α scratch) that requires reading and writing to the connection relationship table in one cycle in which one pixel is input to the temporary labeling circuit, the read is performed in the first cycle, and then the next Write the light in cycles of
The configuration is such that provisional labeling is performed by accessing the connection relationship table once at most every l cycle in which each pixel is input.

[Industrial application field]

In order to clearly identify multiple images that exist independently within each screen that is input at the video rate, the present invention attaches a unique label to each area of each image.
This relates to the pff area separation method.

With the recent improvement in the processing power of computer systems and advances in image processing technology, image processing technology is being used to image the inspection target in parts inspection in factories, blood cell tests in the medical field, cytodiagnosis, etc. Counting the number of internal parts, blood cells, and cells has come to be done.

In this case, it is necessary to identify each of the plurality of objects present in the target image, and so-called labeling is performed for each area that specifies the object.

This labeling process for specific areas in an image is an essential technique for separating multiple objects in an image.
There is a need to speed up the labeling process.

However, when video-rate image processing is performed, the labeling process requires high-speed memory elements used for the connection relationship table required in the label integration process. It is required that the memory device is capable of labeling the video rate.

[Prior art and problems to be solved by the invention] FIG. 4 is a diagram explaining a conventional video rate area separation method, in which (a) shows the configuration of a logical mask, that is, a window, and (b) 1 shows the concept of scanning when processing continuous images, and FIG. 10(c) shows a hardware block diagram.

As shown in FIG. 4 above, examples of the conventional region separation technology based on video rate include "High-speed labeling processor that performs continuous image processing at video rate" 1 by Tetsuro Hattori.
Kento Nonomura, Toshiba Engineering Corporation, Information Processing Society of Japan 35th National Conference (late 1988), so we will omit the details, but the conceptual explanation is as follows.

That is, the continuous image shown in (b) was scanned with a window (referred to as a logical mask in this figure) consisting of 5 pixels ((1 x 3) + (l x 2)) shown in (a) figure. Then, while storing the connection information in the one-dimensional array table memory (LRAM) A when the point P of the window is the pixel value (0/1) of that point and the connection is 8 connections,
Primary labeling is performed based on window patterns divided into eight cases.

At this time, each time scanning of each row is completed, the above-mentioned -dimensional array table memory (LRAM) A is accessed to integrate the label information.

Next, a copy of the -dimensional array to the table memory (LRAM) is made to another memory (LRAM) B, and based on the connection relationship of the provisional label values on the copied table memory (LRAM) B, one frame is This is to integrate M provisional label values into an image (referred to as secondary labeling).

In the high-speed labeling processor developed in this prior art, when each pixel is input, one table memory (L
RAM) A read and table memory (LR) twice.
AM) There may be cases where it is necessary to write A, and in order to process an image consisting of 512 x 512 pixels as shown in the figure at video rate, a memory element that can access the table memory at extremely high speed is required. There is a problem with this.

Further, in this processor, as described above, it is necessary to integrate the labels in the table memory (LRAM) A after scanning each line.

In view of the above-mentioned conventional drawbacks, the present invention provides a video rate area in which a label is attached to each area of each image in order to clearly identify a plurality of images that exist independently within each screen input at the video rate. In the separation method, each time one pixel is input, it is necessary to read once and write twice to the connection relationship table memory, and the connection relationship table memory requires an ultra-high-speed memory element. The purpose is to provide a video rate region separation scheme that solves the problem.

[Means to solve the problem]

FIG. 1 is a diagram showing the principle structure of the video rate area separation method of the present invention.

The above problems are solved by a temporary labeling method configured as follows.

A screen with multiple images that exist independently within one screen is scanned with a window consisting of 5 pixels ((1 x 3) + (I x 2)), and a window pattern is divided into multiple cases. According to the above window, (I
If the first pixel (Z or P) of the window has not been provisionally labeled yet or has a provisional label different from other pixels in the window, the first pixel (Z or P) in the scanning direction or vertical direction, Or window pattern l ~ (41,
(7) to (91) (see Figure 3 (cl)) and the first pixel (P) of the (1x2) window of the window consisting of 5 pixels ((1x3) + (1x2)}) When only is °l, a window pattern ((61) (see FIG. 3 (cl))) that assigns a new temporary label value is used.

At least, the first pixel (I
P) is "1'" and two or more provisional label values are assigned to two or more other pixels, the two provisional label values are compared and the smaller provisional label value is set as above. '1
Window pattern ((1(1) to Q) to be the provisional label value of the above provisionally labeled pixels including the pixels '
(see FIG. 3 (C2)); a frame memory 4 that stores one frame of the provisionally labeled image; and provisional labeling of one frame. After completing the processing, a connection relationship integration circuit 3 integrates the connection relationships of the provisional label values based on the connection relationship table 2, and a connection relationship integration circuit 3 integrates the connection relationships of the provisional label values based on the connection relationship table 2. Correspondingly, a video rate area separation circuit comprising a density conversion circuit 5 for converting the density of each pixel separates the image area, and also separates the image area in one cycle in which one pixel is input to the provisional labeling circuit 1. , When the window pattern (Q[ll-Q91) requires reading and writing to the above connection relationship table 2, the read is performed in the first cycle, and 6 writes are performed in the next cycle, so that each pixel is manually processed. to be done1
The configuration is such that provisional labeling is performed by accessing the connection relationship table 2 at most once per cycle.

[Operation] That is, according to the present invention, in order to clearly identify a plurality of images that exist independently within each screen input at the video rate, a label is attached to each region of each image. In the area separation method, access to a table memory representing the connection relationship of provisional label values given to each pixel is limited to at most once per cycle in which each pixel is input.

FIG. 3 is a diagram showing the labeling logic according to the present invention, in which (a) shows an example of the structure of a connection relationship table, (b) shows an example of the structure of a window, and (cl) and (C2) show the case. (d) shows a method of accessing the connection relation table, and when performing temporary labeling according to the present invention, the above (cl)
, (C2) Provisional labeling is performed based on the case-divided window patterns shown in the figure.

In this figure, L7 represents a label, and R7 represents data at address L in the connection relationship table.

First, the window shown in figure (b) is the same as the window consisting of 5 pixels ((1 x 3) + (1 x 2)) shown in the conventional method, and is shown in this example. Express like.

In the present invention, provisional labeling is performed using window patterns divided into 18 cases as shown in FIGS. (cl) and (C2).

For example, when the window pattern shown in (2) is obtained by scanning the image with the window, nothing is done, but when the window pattern shown in (2) is obtained, the connection shown in FIG. The data R8 obtained by accessing the relation table 2 at the address L is input to the pixel,
be the new temporary label value. (Generally, in the connection relationship table 2, the address Ln (label value) and the data value R,.

Similarly, in the case of window pattern (3), labeling processing is performed to extend the provisional label value in the scanning direction, and window pattern (4 ), labeling processing is performed to extend the temporary label value in the vertical direction. Temporary labeling is also performed for window patterns (7) to (9) using the same logic.

When window pattern (6) is obtained, it is recognized as a new independent area and a new temporary label value of N=N+1 is given to the pixel.

Next, the window pattern α■~(
7) When the temporary label value is obtained, the connection relationship table is read using the provisional label value as the address, and when the relationship R1>L3 is obtained with the data R5, the provisional label value is read.7. Then, the provisional label value of the pixel with the pixel value "1" is set as RI, and the data at address R8 is written to L3 of the connection relationship table 2 to store the connection relationship.

When window patterns Q3) to Q05) are obtained, the temporary label values are compared in size, the smaller label value is set, and the connection relationship is stored in the connection relationship table 2. do.

The above provisional labeling process requires one read access and one write access to the connection relationship table 2.

Also, when the window pattern Q [9 to G barrel is obtained, the connection relationship table is read using the temporary label value as the address, and RI=
When the relationship L3 is obtained, the provisional label value L1. Then, the provisional label of the pixel with the pixel value °1'' is set as L3.

When the final window pattern 09) is obtained,
When reading the connection relationship table using the temporary label value L1 as an address and obtaining data R, generally, R,≦L
1, the provisional label value L1. And pixel value ゛1°
The provisional label value of the pixel is set to R8.

Among the above window patterns, (2) or aω~Q
When pattern 91 appears, the next window pattern is (
1), or (3), or (4).

And (11, (3), (4), (61~(9)
In the pattern (2), it is not necessary to access the connection relationship table 2, but in the pattern (2), it is necessary to read the connection relationship table 2 once, and in the pattern αφ~α, it is necessary to read the connection relationship table 2 once as described above. A read and one write are required.

Therefore, in the present invention, when a sacrificial pattern appears in the aΦ~, the next window pattern is (1), (3), or (4) as described above, and in this window pattern, the connection Focusing on the fact that there is no need to access relational table 2, as shown in figure (d),
(In the cycle of 101 to (1 m pattern, read is performed, and when the pattern (1), (3), or (4) appears next, the necessary write is performed in the pattern of the 0Φ to α field. do.

By controlling in this manner, the connection relationship table 2 only needs to be accessed once at most for any window pattern, and the connection relationship table 2 can be constructed using practical memory elements. Effects can be obtained.

〔Example〕

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

The above-mentioned FIG. 1 is a diagram showing the principle configuration of the video rate area separation method of the present invention, and FIG. 2 is a diagram showing an embodiment of the present invention, where (a) shows an example of the window configuration; b) shows an example of the schematic configuration of the temporary labeling circuit;
The figure shows the labeling logic according to the present invention, and the temporary labeling logic (
(see FIG. 3), and a connection relationship integration circuit 3 that integrates connection relationships on a frame-by-frame basis are necessary means for implementing the present invention. Note that the same reference numerals indicate the same objects throughout the figures.

Hereinafter, the video rate region separation method of the present invention will be explained with reference to FIG. 2 while referring to FIGS. 1 and 3.

In FIG. 2, a window consisting of five pixels ((1×3)+(1×2)l is expressed as shown in FIG. 2(a).

First, when a binarized image input is input to the 5-image window generation circuit 11, the pixel value change circuit 15.
The window pattern consisting of the temporary label value from the pixel value changing circuit 16 is window pattern (1) to α.
The signal is sent to the sound determination circuit 12 and divided into the window patterns shown in FIG. 3 mentioned above.

As is clear from the window pattern of (1) to (41, (61 to α9) shown in Fig. 3, ((1x3) + (1x
2) Based on the provisional label value indicated by the pixel marked ■ in the window consisting of five pixels (corresponding to “Ll” in the window shown in FIG. 3(b)), the related material is generated by the memory address generation unit 1.
3, an address for connection relationship table 2 is generated,
The connection relationship table 2 is accessed by the address, the data (R, etc.) is read out, and fed back to the window pattern (1) to α9 determination circuit 12 to generate the next window pattern.

Based on the generated window pattern and according to the labeling logic shown in FIG. 3, the pixel value changing circuit 15.
and ■ In the pixel value changing circuit 16, ■, ■ pixel (third
The provisional label value of the pixel whose label value has been updated in Figures (cl) and (c2) is determined.

The determined provisional label value is stored in the frame memory 4, and at the time when the provisional label value for one frame is stored, the connection relationship integration circuit 3 reads the frame memory 4 and recognizes the connection relationship. Then, determine an arbitrary label value for the pixels in the mutually connected regions,
The density conversion unit 5 performs density conversion of the pixel corresponding to the label value, thereby completing the final region separation process.

As described above, the present invention uses a video rate area separation method that labels each area of each image in order to clearly identify a plurality of images that exist independently within each screen that is input at the video rate. In, ((1×3)+(1×2)
) A predetermined provisional labeling process is performed based on, for example, 18 window patterns that appear by scanning with a 5-pixel window, and access to the connection relationship table required for one provisional labeling process is performed. If is a two-time pattern of read and write, the next pattern that appears will focus on the fact that there is no access to the connection relationship table, and divide the two memory accesses into two cycles. The feature is that provisional labeling can be performed with a maximum of one memory access per cycle in which one pixel is input.

〔Effect of the invention〕

As described above in detail, the video rate area separation circuit of the present invention has the following functions:
In order to clearly identify multiple independent images, the video rate area separation method attaches a unique label to each area of each image. The screen with the image ((1×3
)+(1×2)) Window pattern divided into multiple cases by scanning with a window consisting of 5 pixels (1)
~(4).

(6) ~A provisional labeling circuit that assigns a provisional label value based on α flaws, a frame memory that stores one frame of the provisionally labeled image, and a provisional A connection relationship integration table that integrates the connection relationships of label values based on the above connection relationship table, and the density of each pixel is converted for the provisional label value stored in the frame memory using the result of integrating the connection relationships. In order to separate image regions in a video rate region separation circuit equipped with a density conversion circuit that performs image processing, it is necessary to read and write the connection relationship table in one cycle in which one pixel is input to the temporary labeling circuit. When using a window pattern (Ql to Q9), reading is performed in the first cycle and writing is performed in the next cycle, and the above connection table is accessed at most once per cycle in which each pixel is input. Since provisional labeling is performed in the following manner, the connection relationship table 2 only needs to be accessed once at most for any window pattern, and the connection relationship table 2 can be constructed using a practical memory element. This makes it possible to economically speed up the labeling process.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing the principle configuration of the video rate area separation method of the present invention. FIG. 2 is a diagram showing an embodiment of the present invention. FIG. 3 is a diagram illustrating labeling logic according to the present invention. FIG. 4 is a diagram illustrating a conventional video rate area separation method. It is. In the drawing, 1 is a temporary labeling circuit. 11 is a 5-pixel window generation circuit. 12 is a window pattern (11 to α odor determination circuit; 13 is a relational memory address generation unit; 15 is ■pixel value changing circuit; 16 is ■pixel value changing circuit; 2 is a connection relationship table; 3 is a connection relationship integration circuit; 4 is a frame memory 5 is a density conversion unit. Ll, Lz, -, R+, - are provisional label values. X, Y, Z, U, I', or ■~■, or L:l+
L2. Ll is a Landau pixel. are shown respectively. Imoto Mokuaki's bisweat 7 and Ittose Rikima 1 Izumi Kata's source Riyo 1 Mr. figure 7
1 Figure ri (Y/) 1) 廿θ The intervention is also ＼D θ0

Claims (1)

[Claims] A screen having a plurality of images each existing independently within one screen is scanned with a window consisting of 5 pixels of {(1×3)+(1×2)}, Connection relation table (2) according to the window pattern divided into multiple cases
In the labeling process that performs provisional labeling by concatenating with 8 connections while accessing the If the pixel has not been provisionally labeled yet or has a provisional label different from other pixels in the window, the provisional label value is given by extending in the scanning direction, vertical direction, or diagonal direction. Window pattern {(1) to (4), (7
) to (9)} and 5 of the above {(1×3)+(1×2)}
When only the first pixel (P) of a (1×2) window consisting of pixels is '1', the window pattern {(6)} that gives a new temporary label value and at least (1× 2) First pixel (P) of window
is '1' and two or more provisional label values are assigned to two or more other pixels, the two provisional label values are compared and the smaller provisional label value is set to '1' above. a temporary labeling circuit (1) that assigns a temporary label value based on a window pattern {(10) to (19)} that is set as a temporary label value of the provisionally labeled pixel including the pixel '; A frame memory (4) that stores labeled images for one frame, and a connection relationship that integrates the connection relationship of provisional label values based on the above connection relationship table (2) after completing the provisional labeling process for one frame. An integration circuit (3) and a density conversion circuit (5) that converts the density of each pixel according to the provisional label value stored in the frame memory (4) based on the result of integrating the connection relationships. The video rate region separation circuit separates the image regions, and one pixel is input to the provisional labeling circuit (1).
Window pattern that requires reading and writing to the above connection relation table (2) in a cycle {(10) to (
19)}, reading is performed in the first cycle and writing is performed in the next cycle, and the above connection relation table (
2) A video rate region separation method characterized by provisional labeling upon access.