JPS62211783A - Picture processor - Google Patents

Abstract

PURPOSE:To separate overlapping pictures without breaking original forms of pictures by performing the length conversion where 1, 2- are added successively as density to results (boundaries of pictures) of contour line tracing of pictures. CONSTITUTION:A length converting means 11 adds 1, 2- as density successively to results of contour line tracing of pictures and decides the presence or the absence of overlapping pictures and the density of boundaries in accordance with the value from a threshold setting means. Pictures having the sholds other than one out of length converted pictures converted by the means 11 are discriminated as overlapping pictures and are converted to the third binary pictures, which are generated by cut parts on boundaries of overlapping pictures, by an exclusive expanding means 13. A classifying means 16 classifies pictures into new pictures which are not overlapping pictures, pictures which cannot be cut, and pictures having cut parts, and they are processed and are outputted as the fifth binary pictures as final results by an OR means 18.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field of the Invention] The present invention relates to an image processing device suitable for separating overlapping particle images.

[Technical Background of the Invention] When measuring the area, circumference, etc. of digital particle images obtained from optical or electron microscopes, it is often found that what should be a plurality of independent particles overlap as shown in Figure 8. If so, it will be necessary to separate them.

As shown in Fig. 9, methods for separating overlapping particles include the first method (opening method) in which the original image of the particles is contracted and then this contracted image is dilated, and the first method is to separate the coordinate string of the figure boundary. The second method, which is performed by numerically calculating the center and radius of the particle from , has been common.

[Problems with Background Art] However, in the first method, as shown in FIG.
There are problems in that parts that should not be separated (cut) (for example, small protrusions) may disappear, resulting in a figure different from the original image, or that separation cannot be performed well if the cutting length is long. Further, the second method has problems in that the processing is complicated and requires a large amount of processing time, and furthermore, it is difficult to form cutting lines.

[Object of the Invention] This invention was made in view of the above circumstances, and its purpose is to separate overlapping figures by forming a cut at the constriction of such a figure, thereby destroying the original shape of the figure. An object of the present invention is to provide an image processing device that can perform image processing at high speed without any problems.

[Summary of the Invention] In the present invention, as shown in FIG. 1, for example, a distance conversion means 11 for obtaining a distance-converted image by distance-converting a first binary image including overlapping figures, and a threshold value k are set. Tall setting means 12 is provided. The setting value of the threshold value setting means 12 is used as a threshold value for dividing the constriction part of the overlapping figure by binarizing the distance conversion image, and m is used as the initial value.

Further, in the present invention, a third binary image is obtained by applying exclusive expansion to the second binary image to avoid fusion of separated figures in the same image and forming a cut portion at the corresponding location. Target expansion means 13, synthesis means 14, and threshold value updating means 15 are provided. Every time a third 21 image is obtained by the exclusive expansion means 13, the synthesis means 14 divides the third 21 images into a smaller value by 1 based on the same image and the threshold set in the threshold value setting means 12.
A fifth 2fa image that includes the above-mentioned cutting part and maintains the original shape of the figure is obtained by combining the fourth binary image obtained by converting the distance transformed image into 21-square with 1-square as the threshold. obtain. The synthesis means 14, in one embodiment, includes a classification means 16, a logical product means 17, and a logical sum means 18. The classification means 16 uses the sixth binary image when the threshold (llk) set in the threshold setting means 12 is m, and uses the previously obtained fifth binary image when it is less than m-1.
By using each fi1 image as a reference image, the fourth binary image is classified into a new image, a cuttable image, and an uncuttable image. The logical product means 17 logically ANDs the cuttable image and the third binary image, and the logical sum means 18 logically combines the logical product result of the logical product means 17, the new image, and the uncuttable image. The sum is taken to obtain the latest fifth binary image.

On the other hand, the threshold updating means 15 determines whether or not the processing can be terminated each time the fifth binary image is obtained by the synthesizing means 14, and if the processing cannot be terminated, the threshold value set in the threshold setting means 12 is changed to 1. Reduce by one. Every time this process of reducing 1 by one is performed, the exclusive expansion means 13 performs exclusive expansion on the fifth binary image as the second 2+III image. The exclusive expansion means 13 also performs exclusive expansion when the distance conversion by the distance conversion means 11 is completed. However, in this case, the sixth binary image obtained by converting the distance-converted image into 2@ using the initial value m, which is the setting value of the threshold value setting means 12 at that time, as a threshold, is the same as the second binary image. Used as a value image. Through the processing of the exclusive expansion means 13, overlapping figures can be separated according to the size of the figure, and furthermore, the original shape of the figure is maintained by the compositing process of the compositing means 14.

[Embodiment of the Invention] FIG. 2 is a block diagram of an image processing apparatus according to an embodiment of the invention. In the figure, 21 is a CPU that controls the entire device, and 22 is a main memory.

A predetermined area of the main memory 22 stores a threshold k (initial 1
1 [m> is set. 23 is an input image memory that stores a binarized input image (for example, a cell tissue plane [gl>) containing overlapping figures. This input image is, for example, an image obtained by binarizing a grayscale (multiple (straight) image). 24 is a distance conversion image memory for storing a distance conversion image of the input image;
5 is an image in which overlapping figures included in the input image are separated;
That is, there is a result image memory that stores an image (result-@O2) having a cut portion (cut line). 26 is the result image 0. Degenerate point image memories 27 to 29 are work memories that store degenerate point images.

31 is a distance transformation processor that performs distance transformation on an input image to obtain a distance transformed image, and 32 is an exclusive expansion processor that performs exclusive expansion on the resultant image Ok to obtain an exclusively expanded image E. This exclusive expansion refers to a process in which separate figures are expanded but not fused, and is performed, for example, by a 1 cycle/4-direction 8-connected exclusive expansion filter. 33 is the image [11E and binary image (binary readout image')
Dk-1 and obtains the result -m o, . In the above image, -1 is the threshold 11 at that time
It is obtained by binarizing the distance-converted image @D using -1 as a threshold value to a smaller value by 1.

CPU 21, memories 23 to 29 and processor 3
1 to 33 are interconnected by an 111111 bus (multipath) 41. Further, the memories 23 to 29 and the processors 31 to 33 are interconnected by an image bus 42.

Next, the operation of an embodiment of the present invention will be described with reference to a flowchart in FIG. 3, a distance conversion explanatory diagram in FIG. 4, and FIGS. 5(a) to (f).
This will be explained with reference to various image examples.

First, the CPU 21 sends the distance conversion processor 31 to the control bus 411! Instruct the distance conversion operation using At this time, the CPU 21 also instructs the final distance of conversion (ji final concentration value). The distance conversion processor 31 is activated by an instruction from the CPU 21, reads the binarized input image from the input image memory 23 via the image bus 42, performs distance conversion,
A distance transformed image is obtained (step 31). This distance-converted image is sent to the distance-converted image memory 2 via the image bus 42.
It is stored in 4.

Now, as shown in FIG. 4, the distance conversion in step S1 is performed on the result of contour tracing of the figure (i.e., from the boundary of the figure).
This is a method of sequentially adding the density (distance 11tla) as 1.2, etc., and is possible using 8-neighborhood, 4-neighborhood, mixed type (octagonal distance conversion by mixing 8-neighborhood and 4-neighborhood), etc. Fourth
The figure shows an octagonal distance transformation image for overlapping circles with a radius of 12.8 (1/4 circle in the figure), and this image (overlapping figure) is divided at a distance (5 in this case) as a boundary. In this case, the figure must be cut according to the procedure described below, and the circle in FIG. 4 indicates the cutting line.

In the distance conversion described above, the distance conversion processor 31 calculates the distance added per pixel (11um<m degrees) from the CPU.
It is checked whether the conversion R final distance specified by 21 has been reached. If reached, distance transform processor 31 ends the distance transform. Distance conversion processor 31
When the distance conversion is finished, the maximum distance wi of the entire distance conversion image (does not necessarily match the true maximum distance)
This is also notified to the CPU 21 (via control 1 < control 1). This maximum distance (density) is 12 in the example of FIG. 4, and becomes 5 if 5 is designated by the CPU 21 as the 314% distance.

Now, the reason for applying the above distance transformation in this example is as follows. The first is that the distance image up to the figure division includes the correct cutting position image. Second
This means that the maximum cutting length specification to avoid unnecessary cutting can be replaced by the final distance specification in the distance conversion operation described above, and this also makes it possible to reduce the number of iterative processes shown in the flowchart in Figure 3. .

Upon receiving the completion notification from the distance conversion processor 31, the CPU 21 sets the maximum distance notified from the processor 31 as the initial value m of the threshold 1i11 in a predetermined @ area of the main memory 22 (step S2). Next c p U
21 is the threshold 1! of the distance transformed image @D for the synthesis processor 33. Instructs smoothing of the binary image by Ik (k-m). As a result, the synthesis processor 33 reads the distance transformed image from the distance transformed image memory 24, binarizes the same image with the threshold value k (-m), and binarizes the same image D (
-D), and by applying a normal km cleaning logic filter to the same image, unnecessary local peaks are corrected, resulting in an image 10k, as shown in FIG. 5(a).
(-0m) is obtained (step S3).

This result image Ok is stored in the result image memory 25 via the image bus 42. At this time, the degenerate point image of the result image OK is stored in the degenerate point image memory 26.

In addition, 21i! The image sensitivity Dk is defined as 1°' among each pixel of the distance converted image (!ilD), which has 11 degrees greater than the threshold value.
, by setting the other pixels to 0''.

When the synthesis processor 33 finishes the process of step $3 described above, it notifies the CPU 21 of this fact.

The CP LJ 21 activates the exclusive expansion processor 32 via the control bus 41 in response to the notification of completion of step S3 from the synthesis processor 33. This causes the exclusive expansion processor 32 to extract the result image Ok(
Here, O) is read out and subjected to exclusive expansion, and FIG.
), obtain the exclusive dilated image E (step 3
4). In this exclusive expansion, unlike a normal expansion operation, the occurrence of fusion accompanying expansion is avoided and a cutting line is formed at the relevant location, as shown in FIG. 5(C). Step S4
The exclusive expanded image E obtained in step 1 is stored in the work memory 27. Note that the number of expansions e in step S4 is 1
．． About two times is enough.

When the above-described step S4 is completed, the exclusive expansion processor 32 notifies the c p U 21 to that effect.

In response to the completion notification from the exclusive expansion processor 32, the CPU 21 instructs the composition processor 33 to perform composition processing using the image @D. This allows the synthesis processor 3
3 is first converted into the distance conversion image memory 24 using the image Dk-1 (i.e., a value 1 smaller than the current threshold k as the threshold).
The image 1g obtained by binarizing the distance conversion surface flilD in
As shown in 1J(d), the images are classified into a new surface (ilN), a cuttable image C, and an uncuttable image ■ (step S5).

This classification involves labeling the binary image Dk-1, checking which label each degenerate point (1 pixel) in the degenerate point image belongs to, and calculating the number of degenerate points included in each label for each label. (that is, obtaining a histogram). Specifically, − to the labeled binary image 0.

The result image is OK in the area corresponding to the shape of the label inside.
If there is no degeneracy point in the image, the figure is determined to be a new image N, if only one exists, it is determined to be an uncuttable image @I, and if two or more exist, it is determined to be a cuttable surface. It is determined as @C.

The new image N, the cuttable surface [11C1, the uncuttable image @I, is stored in the work memory 28.

After completing step S5, the synthesis processor 33 applies a smoothing logic filter to the new image N in the work memory 28 to create an image N in which unnecessary local peaks have been corrected, as shown in FIG. 5(e). ′ (step S6
). This image N' is stored in the work memory 29. Next, the synthesis processor 33 performs a logical product of the cuttable surface IIC in the work memory 28 and the exclusive expanded image E in the work memory 27 to obtain an image A that maintains its original shape and has a cutting line (step 87). . This picture mA is the work memory 29
is stored in Next, the synthesis processor 33 calculates the logical sum of the image A in the work memory 29, the image N- in the same memory 29, and the image I in the work memory 28, as shown in FIG. 5(f).
The result images 0 to 1 shown in (step S8) are obtained. This result image @O,-1 is stored in the result image memory 25.

In other words, go to the previous result image or go to the latest result image i1.

can be rewritten as

When the synthesis processor 33 finishes the process of step S8 described above, it notifies the CP tJ 21 to that effect.

Upon receiving this termination notification, the CPU 21 determines whether the threshold value k set in a predetermined area of the main memory 22 has reached 2 (step S9). If step S
If the determination result in step 9 is No, that is, if k is 3 or more, the CPU 21 sets −1 to new k and executes step 5.
IO), L. After that, the exclusive expansion processor 32 is activated to perform the process of step S4 described above. Then, when the above processing is repeated and k becomes 2, CPU2
1 determines the end of a series of processes. At this time, the result image memory 25 stores a result image 01 in which the overlapping figures are separated from each other (that is, with cutting lines) while maintaining the original form of the binary image including the overlapping figures in the input image memory 23. .

As described above, according to the apparatus shown in FIG. 2, overlapping particles (figures) necessary for analysis of particles m (figures) of cellular tissues and metal tissues can be separated. Also, in the above process, input 2
By reversing 1'' and 0'° of the value image, the device shown in Figure 2 can automatically connect the missing parts (inside the dotted lines) of the thinned image as shown in Figure 6, and furthermore, the device shown in Figure 7 can As shown in the figure, it can be applied to automatic connection of scattered particle groups.

Note that by configuring the image bus 42 in FIG. 2 with a plurality of image data buses and operating the distance conversion processor 31, exclusive expansion processor 32, and synthesis processor 33 in parallel, the image bus 42 shown in FIG. Processing can be performed at high processing speed. Also, instead of the binarized input image,
It is also possible to use a grayscale image *<multivalued image). In this case, the distance conversion processor 31 can be made unnecessary by using a grayscale image instead of the distance conversion plane tilD.

[Effects of the Invention] As detailed above, according to the present invention, overlapping figures can be separated at high speed without destroying the original shape of the 11 figures by forming a cutting part at the constriction of this type of figure. I can do it. Further, when determining the cutting portion, the maximum cutting length can be indirectly specified by specifying the distance 11t (If! degrees).

[Brief explanation of drawings]

FIG. 1 is a block configuration diagram for explaining the present invention, FIG. 2 is a block configuration diagram of an image processing apparatus according to an embodiment of the present invention, FIG. 3 is a flowchart for explaining the operation, and FIG. 4 Figures 5(a) to 5(a) are diagrams explaining distance conversion.
f) is a diagram showing examples of various images generated according to the flowchart in FIG. 3, FIG. 6 is a diagram showing a case where the present invention is applied to automatic connection of missing parts of thinned images, and FIG. 7 is a diagram showing examples of images generated according to the flowchart of FIG. FIGS. 8 to 10 are diagrams illustrating the case in which the method is applied to automatic connection of scattered particle groups, and are diagrams explaining problems in the conventional method. 21...CPU, 23...Input image memory, 24.
...Distance conversion image memory, 25...Result image memory,
26... Degenerate point image memory, 27-29... Work memory, 31... Distance conversion processor, 32... Exclusive expansion processor, 33... Synthesis processor. Applicant's representative Patent attorney Takehiko Suzue Figure 4 Figure 5 rM (Part 1) (CI) (e) (f) Figure 5 (Part 2) Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 figure

Claims (8)

[Claims] (1) A distance conversion means that performs distance conversion on a first binary image including overlapping figures to obtain a distance converted image, and an initial value m is initially set as a threshold value for binarizing the distance converted image. a threshold value setting means for obtaining a third binary image by applying exclusive expansion to the second binary image to avoid fusion of separated figures in the same image and forming a cut portion at a corresponding location; by means of exclusive expansion means and the third expansion means mentioned above.
Each time a binary image of compositing means for composing the fifth binary image including the cut portion with the value image;
and threshold updating means that determines whether or not the processing can be completed each time a value image is obtained, and if the processing cannot be completed, reduces the threshold value k set in the threshold value setting means by one, and the exclusive expansion means includes: When the distance conversion by the distance conversion means is completed, the sixth binary image obtained by binarizing the distance conversion image using the initial value m set in the threshold value setting means as a threshold is The fifth binary image is subjected to exclusive expansion as the second binary image, and the fifth binary image is subjected to exclusive expansion as the second binary image each time the threshold setting means is updated by the threshold value updating means. An image processing device characterized by: (2) When the threshold k set in the threshold setting means is m, the synthesizing means converts the sixth binary image into the sixth binary image when the threshold k set in the threshold setting means is m-1 or less. In the case of , the classification means classifies the fourth binary image into a new image, a cuttable image, and an uncuttable image by using the previously obtained fifth binary image as a reference image, respectively; a logical product means for logically multiplying the cuttable image and the third binary image, and logically ORing the logical product result of the logical product means, the new image, and the uncuttable image to obtain the latest fifth binary image. 2. The image processing apparatus according to claim 1, further comprising a logical sum means for obtaining a binary image. (3) The image processing device according to claim 2, wherein the sixth binary image and the new image are smoothed. (4) The image processing apparatus according to claim 1, wherein the threshold value updating means determines the end of the processing when detecting that the threshold value k set in the threshold value setting means is 2. (5) a threshold value setting means in which an initial value m is initially set as a threshold value for binarizing a multivalued image including overlapping figures;
Exclusive expansion means for obtaining a second binary image by performing exclusive expansion on the first binary image to avoid fusion of separated figures in the same image and form a cut section at a corresponding location; Every time the second binary image is obtained by the exclusive expansion means, the same image and the multivalued image are binarized using a value k-1 which is one smaller than the threshold value k set in the threshold value setting means as a threshold. a synthesizing means for synthesizing the third binary image obtained by the above to obtain a fourth binary image including the cut portion; and threshold updating means that determines whether or not the process is possible and decreases the threshold value set in the threshold setting means by one if the termination is not possible, and the exclusive expansion means sets the threshold value set in the threshold setting means in an initial state. Exclusive expansion is performed using a fifth binary image obtained by binarizing the multivalued image using the initial value m as a threshold as the first binary image,
An image processing device characterized in that the fourth binary image is subjected to exclusive expansion as the first binary image each time the threshold value updating unit updates the set value of the threshold value setting unit. (6) When the threshold k set in the threshold setting means is m, the synthesizing means converts the fifth binary image into the fifth binary image when the threshold k set in the threshold setting means is m-1 or less. a classification means for classifying the third binary image into a new image, a cuttable image, and an uncuttable image by using the fourth previously obtained binary image as a reference image; a logical product means for logically multiplying the cuttable image and the second binary image, and logically ORing the logical product of the logical product means, the new image, and the uncuttable image to obtain the latest fourth binary image. 6. The image processing apparatus according to claim 5, further comprising a logical sum means for obtaining a binary image. (7) The image processing device according to claim 6, wherein the fifth binary image and the new image are smoothed. (8) The image processing apparatus according to claim 5, wherein the threshold value updating means determines the end of the processing when detecting that the threshold value k set in the threshold value setting means is 2.