JPS5953968A - Picture processing device - Google Patents

Abstract

PURPOSE:To measure and recognize the pattern of a blood corposcle image and an industrial product, etc. at a high speed, by switching alternately two tag memories at each area conversion, and executing an area compensation by a feature quantity. CONSTITUTION:The tag memory 61 stores a type number and a tag number to which each picture element of a picture belongs. Also, contents are written from a CPU8, and in accordance with the type number signal of each picture element outputted from the memory 61, a tag number signal and a bit pattern output from an operating part 4, an area converting table 51 outputs a new type number signal and a bit pattern signal. In accordance with the output of this table 51, the tag output of a numbering circuit 52 for giving a number, and a type number output from the corresponding table 51 are stored in the tag memory 62. Subsequently, the memories 61, 62 are switched alternately at every area conversion, and the area compensation by a feature quantity is executed.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an improvement in an image processing device that captures an object with a TV camera and performs image measurement of the object from the image.

When capturing an object with a color TV camera and performing image measurement of the object from that image, the object to be measured (e.g. red blood cells, white blood cells, nucleus, cytoplasm, etc.) from the image (e.g. blood image) is Accurate separation is the most basic and important thing.

To separate the target object area, use the spectral characteristics of the target object, and use light of a wavelength where the absorption of the target object is sufficiently larger than that of other objects. There is also (U, S Patent 3.999,047
In this case, a special optical filter is required, a general-purpose color TV camera cannot be used, and the S/N ratio is poor. Normally, an image that has been subjected to color linear calculations is binarized using an appropriate threshold value to divide the area into areas, but with threshold processing alone, areas with similar colors are also cut out at the same time, making it difficult to accurately divide the area. Difficult to separate.

The present invention has been made in order to eliminate the above-mentioned drawbacks, and provides an image processing device that can process gray scale images obtained by a TV camera and accurately separate regions of objects included in the image. The purpose is to

According to the present invention, the above object can be achieved by performing area correction by a small amount of t1 on a binary image separated by calculation from an image obtained by a TV camera.

The present invention will be fully explained below with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of an image processing apparatus according to the present invention. FIG. 2 is a block diagram illustrating a main part configuration for explaining a part of the operation of the 1dMI diagram. 1 is a system bus used for exchanging signals between each part of the image processing device; 2 is connected to the system bus 1; and 2 is an image signal from a TV camera (in the case of this embodiment, a general-purpose color TV camera); A/D conversion unit that digitizes every pixel of the VD,
3 is an image memory unit connected to the system bus 1 and inputs and stores digital image signals from the A/D converter 2; 4 is connected to the system bus 1 and receives digital image signals from the image memory unit 3; A calculation unit 7 inputs the digital image signal and performs color calculation and binarization for each pixel, and 7 is connected to the system bus 1vC, counts up the contents of the addressed counter, and calculates the feature amount of each area. 8 is an addressable counter that performs counting and creates a color calculation histogram; 8 is a processor section that is connected to the system bus 1 and sets parameters and sends and receives data to each section; 9 is a processor section that sends a clock signal to each section; Reference numeral 10 denotes a clock generator which supplies a domain converter, the details of which are shown in FIG. Reference numeral 6 denotes a tag memory unit connected to the system bus 1, which stores a type notification to which each pixel of the color image belongs (hereinafter referred to as a notification indicating the type of area to which a pixel belongs) and a tag notification (A tag attached to each area). , f (hereinafter referred to as)) 61 and 62.
It is composed of 5 is a tagging section connected to the system bus 1; In this tagging section 5, the contents are written in by the table write signal TW from the processor 8, and the type signal TPI for each pixel output from the tag memory 61 and the tag j16.
An area conversion table 52 outputs a new type A signal TP2 and a new bit pattern output BI)2 in response to the TGI and the bit pattern output BPI from the arithmetic unit 4; 52 is a bit from the area conversion table 51;・This is a numbering circuit that performs numbering corresponding to pattern output BP2. The type input signal TP2 from the area conversion table 51 and the tag point output TG2 from the numbering circuit 52 are added to the tag memory 62 and stored therein. The two tag memories 61 and 61j: their roles are exchanged each time an area is converted. Fig. 2 177b1,
jlfy, 2 tag memories respectively operate as tag memories 61 and 62.

The operation of this embodiment will be explained below along with the case where a blood cell image is processed. The blood cell image obtained by the color TV camera is added to the A/D converter 2 as an image signal VD, and the A/D
The digital data after conversion is R (red), G (green), B (
Blue] is written to the image memory 3 as three-color data. As shown in FIG. 3, the data stored in the image memory 3 consists of a white blood cell area WC1, a red blood cell area RC, a nuclear area N, a cytoplasmic area C9, and a background area BG. Next, determine the optimal color calculation coefficient (r1+91+bl) for separating the background region BG and the blood cell region (red blood cell region RC, white blood cell region WC).
is set in the calculation unit 4 by the processor 8. The contents of the image memory 3 are sequentially read out to the arithmetic unit 4 pixel by pixel to calculate R, G.
, B C linear introduction (color 3 contract) e, =r, ・R-
Find the value of 11,・c+b,-I3, and obtain the calculation result e,
The addressable counter 7 is sequentially counted up as the f address, and a color calculation histogram (frequency distribution of color calculation values e1 (pixels)) is formed in the addressable counter 7. FIG. 4 is a color calculation histogram for the image data of FIG. The processor 8 reads the contents of the addressable counter 7 (color calculation histogram) and determines a threshold value T1 suitable for separating the background region BG and the blood cell region. threshold T,
It is sent from the fd processor 8 to the arithmetic unit 4. The contents of the image memory 3 are read out to the arithmetic unit 4 pixel by pixel again.
==e, 'r, == r,' R+ 91' G
+ b,' B T.

The bit pattern output BPI, which becomes a "c binary signal" depending on the sign of the value of C4, is sent to the area conversion table 51. For the pixel C1〉0, the type signal output TP2 is type yffi O, C, (O
The processor 8 writes the table (data is written by iMTW to %t=e) so that the pixels in the background area are of type A3.As a result, the pixels in the background area are of type A3 in the tag memory 62<Aε2 tag memory , the pixels in the blood cell area are scaled to an appropriate size as type A3. Background area A in Figure 5
50 and the tag memory 62 (contents of the tag memory 2) separated from the blood cell area A53&C.

Next, blood cell region A53 'f What is the optimal color calculation coefficient (r2+f2+b) for separating into white blood cell region and red blood cell region?
<Set from processor 8 to calculation unit 4. In the same manner as when separating the background region and the blood cell region, a color calculation histogram is entirely formed in the addressable color, and the processor 8 sets the threshold T2 for separating the white blood cell region and the red blood cell region R.
is calculated and set in the calculation unit 4. The contents of the image memory 3 are read out again pixel by pixel to the calculation unit 4, and C2=C2-T2=r2・R+s'2・G+b2
・B-T.

The bit pattern output BPI, which is a binary signal corresponding to the positive or negative value of C2, is sequentially sent to the area conversion table 51. The area conversion table 51 is preliminarily filled with new data by the table convolution signal TW of the processor 8, and when the type signal TPI is 0 for each type (background area A50), the type shop signal output TP2 is Type flat 0. When the Taigu signal human power TPI is type 3 (nnn area A33 and C2〉0), the type signal output TP2 becomes type 23, and when the type signal human power TPI is type A3 and C2〈0, the type signal is Output TP2 becomes type 1.

The roles of the two tag memories are also reversed, with the point 2 tag memory serving as the tag memory 61, and the point 2 tag memory serving as the tag memory 62.

Therefore, the contents of the tag memory 61 include the background area A50.
A separated pattern is written in the blood cell area A53, and this content is the area conversion table 51 of the above content.
62 (A
1 tag memory), pixels in the background area are of type A O,
Pixels in the white blood cell region are type A3, and pixels in the red blood cell region are type B1. Figure 6 shows the background area A60.
This is a pattern display of the contents of the tag memory 62 (A1 tag memory) which is separated into white blood cell area A63 and red blood cell area A61. When converting the above white blood cell region and red blood cell region separation, region conversion table 5
1 outputs a bit pattern output BP2 at the same time as a type communication signal output TP2. The bit pattern output BP2 at this time is 1 only when the type signal output TP2 is type A, 1 (red blood cell area A61), and 0 in other cases. After inputting and numbering each area, this tag output TG2 is written in the tag memory 62 (AI tag memory) in correspondence with the type of the same pixel and is sized appropriately. FIG. 7 shows the contents of the tag memory 62 numbered in the red blood cell area A61.

The above tag output TG2U is added to the address counter at the same time, and the number of pixels (ff1J
Count the number of stones. The browser 8 reads the content of the addressable counter 7 and finds tag points for areas whose area is smaller than a certain value.

Then the roles of the two tag memories are swapped, horse 1,
Each A62 tag memory has a tag memory of 61.62.
becomes. In addition, in the area conversion table 51, when the type signal human input TPI is type A, and the tag input TGI is a tag point with a small area, the type signal output TP2 is type A.
3, and when the type signal TP1 is type no. 1 and the tag report input TGI is not a small area tag point, the type signal output TP2 is type I61. l! When the type A signal manual TPI is type AO and type A3, new data is written from the grosser 8 so that the output signal TP2 becomes the tenon type signal output TP2. In this state, if the tag memory 61 is separated into the background area A60, the white blood cell area A63, and the red blood cell area A61 (,%1), if the contents of the tag memory are output to the area conversion table 51 for each pixel, the area will be small. The red blood cell area is changed to a white blood cell area and input into the tag memory 62 (red blood cell area correction).

FIG. 8 shows the tag memory 62 (A,
2 tag memory) is displayed in a pattern (As2...background area, 881...red blood cell area,
A2B...white blood cell area). (There are some parts where the green red blood cell area has been changed to a white blood cell area, but this will be corrected in the next white blood cell area correction process.) When changing the area for red blood cell area correction above, Conversion table 51
From then on, a bit pattern output BP2 is applied to the numbered circuit 52 in synchronization with the type signal output TP2. At this time, the bit pattern output BP2U and the type signal output TP2 become 1 only when type number 3 (83 to the white blood cell area), so the numbering circuit 52 numbers the white blood cell area A83. Numbered circuit 52
Tag output from TG2+/;j: Tag memory 62
(A, 2 tag memory) is written in correspondence with the type of the same pixel. No. 9 [4] shows the contents of the tag memory 62 (A, 2 tag memory) in which the white blood cell area A83 is numbered. At the same time as the above tag notification output TG2 fl, the addressable counter 7 is counted up using the tag point as an address, and the area (number of pixels) of each region of the white blood cell region A83 is determined within the addressable counter. Processor 8 reads this addressable call h) content sound,
The tag j6 of the (white blood cell) region with the largest area is determined.

Next, the roles of the two tag memories are discussed again in the literature, and A.I.
, J102 tag memory is each tag memory 6
It becomes 2.61. The area conversion table 51 also shows the type signal output TP when the type signal TPI is type 116.3 and the tag input TGI is the tag point with the largest area.
2 becomes type notification 3, type signal input TPI is type notification 3, and tag j6 human power TGI is not the largest area tag shop, type signal output TP2 becomes type 116.1, type signal input TPI becomes type 0. and type A, and in the case of 1, the kneading type signal output TP2
New data is written from the processor 8 so that. In this state, the tag/memo after red blood cell area correction') 6
1 (A2 tag memory) is sequentially output to the area conversion table 51 pixel by pixel, the white blood cell areas other than the white blood cell area with the largest area are changed to red blood cell areas and are converted into tag memo by the typewriter signal output TP2. 'J 62 (11
(y, 1 tag memory) (white blood cell area correction).

Figure 10 shows the tag memo after the above white blood cell area correction.
62 (162 tag memory) is displayed in the self-answer pattern (A100...background area, A:l-01
... Red blood cell area, AlO3... White blood cell area. During the area conversion for white blood cell area correction, the type signal output TP2 from the area conversion table 51 is simultaneously applied to the addressable counter 7 as an address signal.

The processor 8 sends the calculation unit 4 in advance a color calculation coefficient (r3.
V3. b3) is set. In synchronization with the area conversion for white blood cell area correction, the contents of the image memory 3 are read out to the calculation unit 4, and the calculation output e5 from the calculation unit 4 is sent to the addressable counter 7 as the other address signal.
added to. As a result, the addressable counter 7 is addressed with the color calculation result e3 and the white blood cell area corrected type and counts up, forming a color calculation histogram in the addressable counter. The processor 8 selects type 46.3 (color calculation histogram) of the contents of the addressable counter 7 (
The part corresponding to the white blood cell area A103)K is read out, and the optimum threshold value T5 for separating the nuclear area and the cytoplasmic area is determined. The threshold value T3 is sent from the processor 8 to the arithmetic unit 4.

Hereinafter, the nuclear region is separated and corrected from the white blood cell region, and the operating procedure at this time is the same as the case where the red blood cell region is separated and corrected from the blood cell region described in IIJ. Finally, the nuclear area is type A3, the cytoplasmic area is type 2, the red blood cell area is type 1, and the background area is 1: A, 0.
The content shown in Figure 1 is A, 1 tag memory (tag memory).
(AIIO... background area, A111... red blood cell area, A112... cytoplasmic area, A113... nuclear area).

FIG. 12 compares the region separation result (5) based only on color calculation and the result (B) with region correction added for another image, and the effect of region correction is clear.

In the above embodiment, area is used as the feature amount used for region correction, but other feature amounts such as circumference, eye roughness, etc. can also be used.

Further, an external computer may be used as the processor 8 in the above embodiment. In this case, the processor 8 takes charge of the communication function with the external computer.

In addition, in the above embodiment, the color calculation coefficient can be freely set according to the color of the object, and the optimum value for each color image is determined by the processor to perform region separation, so it is not affected by conditions such as staining and light source. Regions can be separated accurately without being affected.

Furthermore, since the present invention can use a general-purpose color TV camera and is fast, it can be applied not only to blood cell images but also to pattern measurement and recognition of general industrial products.

Further, in the above embodiment, the case where a color TV camera is used has been described, but the present invention is not limited to this, and can be similarly used for black-and-white rice images. In this case, the arithmetic unit only needs to be a binarization circuit, and the configuration of the image memory can be simplified.As described above, according to the present invention, a grayscale image obtained by a TV camera is processed, Accordingly, it is possible to realize a high-speed image processing device that can accurately separate regions of a target object.

[Brief explanation of drawings]

FIG. 1 is a configuration block diagram showing one embodiment of the present invention, and FIG.
The figure is a block diagram of the main parts of a part of figure 1, Figure 3 is a color gradation image stored in the image memory,
The figure is a color calculation histogram for the image data in Figure 3, Figure 5 is an image in which the gray scale image in Figure 3 is separated into a background area and a blood cell area, and Figure 6 is an image of the blood cell area in the image in Figure 5 with red blood cells. Image separated into a region and a white blood cell region, 7th
The figure shows an image in which the red blood cell areas in Figure 6 are numbered, and Figure 8 t.
ri An image obtained by correcting the red blood cell area on the image in Fig. 6, Fig. 9 is an image in which the white blood cell areas in Fig. 8 are numbered,
Figure 10 is an image obtained by correcting the white blood cell area on the image in Figure 8, Figure 11 is an image obtained by separating and correcting the nuclear area in the white blood cell area of the image in Figure 10, and Figure 12 (A
) is an image in which regions have been separated only by color calculation, and FIG. 12(B) is an image in which region correction has been added. 1... System bus, 2... A/D variable section, 3
. . . image memory section, 4 . . . arithmetic unit, 5 . Processor section, 51... area conversion table, 52... numbered circuit, 61.62...
Tag/memory cuff-1 value e1 rhinoceros 5 figure tail 6 figure M7 figure 8 leather figure 9 figure 10

Claims (1)

[Claims] (1+Processing an image signal obtained by a TV camera,
Separate the region of the object included in the image and identify the object's features'
An image processing device that performs t k measurement, and is characterized in that it performs υ region correction using feature amounts. +21. In an image processing device that processes an image obtained by a TV camera, separates the region of an object included in the image, and measures the feature amount of the object, the type and tag to which each pixel of the image belongs is stored. A first tag memory, the contents of which are written by the processor, corresponds to the type signal and tag signal for each pixel output from the first tag memory, and the pit pattern output from the calculation section. A region conversion table that outputs a new type signal and a pit pattern signal, a numbered circuit that performs numbering corresponding to the pit pattern output from this region conversion table, and a numbered circuit that performs numbering corresponding to the pit pattern output from this region conversion table, and tag notification output and the corresponding type j from the area conversion table
and a second tag memory that inputs and stores six outputs, and performs area correction based on feature amounts by alternately switching between the two tag memories for each area conversion. . (3) An image processing device V that processes images obtained by a TV camera, separates the regions of the object included in the image, and measures the feature amounts of the object, is used for sending and receiving signals between each part. a system bus connected to the system bus, an A/D conversion unit connected to the system bus and digitizing the signal from the TV camera pixel by pixel;
An image memory unit that inputs and stores the digital image signal from the A/D conversion unit is connected to the rJ11 Sysna system, inputs the digital image signal from the image memory unit, and performs calculations for each pixel. The arithmetic unit that performs all the operations and the system
an addressable counter that is connected to the system bus and counts up the contents of the addressed counter to count the feature amount of each area and create a calculation histogram; A tag memory section for storing a tag Af is connected to the system bus, and a bit number output for each pixel from the arithmetic section and a type information for each pixel from the tag memory section are connected to the system bus. It connects to the tagging section, which outputs a new type signal and a new tag signal in response to the Ig and tag A signals, and the oil shift system, and sends and receives parameter sets and data to each of the above sections. What is claimed is: 1. An image processing device comprising: a processor section, and performing all region correction based on feature amounts.