JPH07210671A - Image analysis method - Google Patents

Abstract

PURPOSE:To reduce the burden to an analyzer by using the density difference between an arbitrary image and another image to determine the threshold of another image based on the threshold which the analyzer determines for the arbitrary image. CONSTITUTION:An image processing processor 7 performs variable density processing, binarizing processing, image analysis, etc., of inputted image data, and variable density image data is stored in a variable density image memory 8, and binary image data is stored in a binarizing memory 9. When a digital image on a printed matter or the like including the analysis object image is displayed on a monitor screen and is binarized in accordance with the variable density by the threshold and is subjected to binarization, the threshold is determined based on the reference density in an arbitrary first image, and the highest frequency value (Fmax-1) of density distribution of the image is measured, and differences between this highest frequency value (Fmax-1) and highest frequency values (Fmax-N) of density distribution of second and following images are added to respective picture elements of second and following images at the time of binarizing second and following images, and thereafter, analysis is performed based on the threshold.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is an image analysis method for an image such as a printed matter or a photograph, or an enlarged image by a microscope or the like, and an accurate image analysis can be performed without necessarily requiring skill. It is a possible method.

[0002]

2. Description of the Related Art An image is displayed on a monitor screen for displaying an image, and the image is binarized according to the lightness and shade, that is, the gray level. For example, the area and length of a high density (low gray level) portion can be measured. Has been done. As an example of these methods, a digital camera or A
There is a method of storing a D / D converted image in a frame memory and determining a threshold value (THR) for extracting an image to be analyzed as a binarized image from an image displayed on a CRT monitor. In a series of multiple samples, the THR of the first sample (target image) was determined by the analyst to visually determine the standard image density in the image on the CRT A density level when a portion having is selected as a target image and a binarized image is overlaid on the image without any excess or deficiency is defined as THR. In other words, the analyst visually determines the threshold density for the target image on the monitor screen, compares the binary image binarized by this threshold with the original image, and determines the area and length of the target portion. 2 suitable for measuring
A method of determining a threshold value which becomes a value image by trial and error has been used. Further, JP-A-60-100032 and JP-A-2-232550 disclose that an image density distribution area histogram is created and the threshold value is set using the bottom of the histogram.

[0003]

Even if the same sample or the same series are used, the reflectance of the sample itself, even if the external conditions (optical illumination brightness, TV camera condition, etc.) are constant in a plurality of samples and fields of view, Since the transmittance is not always constant,
Since the density (gray level) of the image that serves as a criterion for determination and the density of the target image vary, the target image is
It cannot be covered with a binary image without excess or deficiency.

Therefore, the T
You have to decide HR. Even if it is possible to set the optimal THR for each sample, the density of the reference image is different in this work, and because of human sensory processing, the optimum value at that time is set, but it is not a little wide. The optimum THR has a value of "3", and it is difficult to set a constant THR that is optimum for each image. Therefore, variations in analysis results and the reliability of analysis are impaired. Skill will reduce the variation and increase the reliability, but it will be a time-consuming and troublesome task for the analyst. If there are more than one analyst, the error becomes even larger.

If the density distribution of the target image rises sharply, it is not so difficult, but
Normally, the image has a density gradient, and it is difficult to set the boundary with the reference image. In other words, in these methods, when the target image is extracted and the length and area are measured by the threshold value with a constant gray level, fluctuations in illumination brightness and image reflectance and transmittance are irrespective of the measurer. Due to the fluctuation of, the analysis result may deviate. That is, if the brightness of the image including the analysis target is too low or too high, the brightness of the reference image fluctuates and TH
This is a cause of increasing the difficulty of setting R, making measurement difficult, and lowering the reliability of analysis results.

In the method of binarizing the gray level with a threshold value which is a predetermined value as described above, there has been a demand for a method of performing a reliable image analysis with no individual difference. Even if it is difficult to set the threshold value based on the bottom of the density distribution histogram, the present invention determines the threshold value after the analyst determines the threshold value for an arbitrary image when determining the threshold value for a plurality of similar images. An image analysis method that reduces the burden on the analyst by determining the threshold value of another image by using the difference in density between that image and another image based on the value, and reduces the variation in threshold value determination The purpose is to provide.

[0007]

The present invention provides a method for performing image analysis by displaying a digital image of a printed matter or the like including an image to be analyzed on a monitor screen, binarizing the digital image with a threshold value according to shading, and performing image analysis. , A threshold value is determined based on a reference density in an arbitrary first image, the mode (Fmax-1) of the density distribution of the image is measured, and the mode (Fmax-1) And the mode value (Fmax-N) of the density distribution of the second and subsequent images are added to each pixel of the second and subsequent images in the binarization of the second and subsequent images, and It is an image analysis method that analyzes based on a threshold value.

The image is an image over the entire surface of the monitor, and the target image is also included in the image. Further, the target image means an image to be binarized and measured (analyzed). Furthermore, in addition, it goes without saying that not only the case of plus (+) but also the case of minus (-) is possible. If the difference is not added to the second and subsequent images, TH
The same effect can be obtained by adding the difference to R.

In a series of images, the THR of an arbitrary first image (reference image) is determined by an analyst by visually deciding a reference image density in the image on the CRT and comparing it with the image. An image having a certain density or more is selected as a target image, and a density level when the binarized image is overlaid on the image without any excess or deficiency is used as a threshold value (THR). Next, in the analysis of the second and subsequent target images, the THR is set to be the same as that of the first image, and the measured mode of the density distribution of the first image (Fmax-1) and the second and subsequent target images are calculated. In the binarization of the second target image, the difference from the mode value (Fmax-N) of the density distribution is added to the densities of the second and subsequent images, that is, each pixel density, and then based on the threshold value. It is a method of analysis. Of course, the third and subsequent ones can be analyzed similarly.

According to the method of the present invention, in the analysis of a plurality of images, the mode (Fmax-1) of the density distribution of the first image whose THR is determined and the mode of the density distribution of the second and subsequent images. The difference from the value (Fmax−N) is added to each pixel of the second and subsequent images including the analysis target, and Fmax−
By setting N to be equal to Fmax-1, it is possible to analyze the second and subsequent images with the same THR to extract the analysis target images without excess or deficiency, that is, to perform analysis efficiently and without skill.

In order to obtain a proper binary image at a constant THR by taking in a color image with an RGB color TV camera,
This is possible by making the Fmax of each of RGB the same as the first Fmax.

In addition, in the image analysis method, a histogram of the grayscale distribution area is created according to preset grayscale levels of a plurality of similar grayscale images, and the threshold value of the corresponding image is determined based on the histogram. The threshold value of the image is determined, and the difference obtained by subtracting the density of the predetermined peak of the histogram representing the feature of another image from the density of the predetermined peak of the histogram representing the feature of this arbitrary image is calculated.
The density obtained by subtracting the difference from the threshold value of an arbitrary image is set as the threshold value of another image.

The density of the predetermined peak is the density representing the mode of the density distribution.

Further, the density of the predetermined peak is a density representing the second largest peak value after the mode of the density distribution.

Further, the density of the predetermined peak is assumed to be the density of the peak near the density representing the background of the image.

Comparing the histograms of the grayscale distribution areas of an arbitrary image and another image with respect to similar images, since the two images are similar, the shape of the histogram, that is, the relative position or shape at which the peak occurs becomes similar. . Of these peaks, the density of the peak of the histogram showing the features common to both images is obtained, and the distance between the position of the peak of the histogram showing the features of the arbitrary image and the threshold of the arbitrary image is obtained. Next, the threshold value of the other image is determined from the density of the peak of the histogram representing the feature of the other image so that the distances are the same. As a result, the thresholds of other images can be set on the same basis as when the thresholds of arbitrary images are set.

Therefore, the difference obtained by subtracting the density of the peak of the histogram representing the feature of another image from the density of the peak of the histogram representing the feature of the arbitrary image is obtained, and this difference is subtracted from the threshold value of the arbitrary image. By setting the density as the threshold value of another image, it is possible to make the distance between the threshold value and the density of the peak representing the feature the same in an arbitrary image and another image similar thereto. If the difference is negative,
Subtracting means adding the absolute values of the differences.

By using the density representing the mode of the density distribution as the density of the predetermined peak, the density of any image and other images can be made substantially the same.

Further, by using the density representing the next largest peak value after the mode of the density distribution as the density of the predetermined peak, the threshold value which makes the best use of the characteristics of the next peak value can be determined.

Further, by setting the density of the predetermined peak as the density of the background (background) of the image, even if the area of the background changes greatly, the threshold values of the arbitrary image and other images can be The distance to the density of the ground can be the same.

[0021]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing the arrangement of an apparatus for realizing this embodiment. An image pickup device 15 is attached to the microscope 1. The stage 16 on which the measurement sample is placed operates the plane moving mechanism 17 in response to a signal from the auto stage driver 10, and the pulse motor provided in the stand adjusts the plane position to move in the XY direction. The vertical moving mechanism 18 is operated by 11 to move the stage 16 in the vertical direction.

The A / D converter 2 converts the input data from the image pickup device 15 from analog to digital, and the input buffer 3 temporarily stores the digital data. The bus 4 transmits signals, the program memory 5 stores a program that defines the operation of the apparatus, and the CPU 6 controls the entire apparatus according to this program.

The image processor 7 performs grayscale processing, binarization processing, image analysis, etc. of the input image data, the grayscale image memory 8 stores the grayscale image data, and the binarization memory 9
Stores binary image data. In response to an instruction from the CPU 6, the auto stage driver 10 controls the plane moving mechanism 17 to move the stage 16 on which the measurement sample is placed in the X and Y directions, and sets the measurement position and area of the measurement sample. The autofocus driver 11 receives a vertical control command from the CPU 6 and controls the vertical movement mechanism 18. The output buffer 12 temporarily stores the data to be output, the D / A converter 13 converts the output data from digital to analog, and the CRT 14 displays the output data on the screen.

When measuring a tissue such as a metal tissue or a living tissue, a large number of samples are often taken out from the same sample or the same series of samples, and many similar tissues are often inspected. Image)), the analyst sets a threshold value while looking at the screen, and for other images, the difference from the first image is displayed with an appropriate parameter such as brightness, and the first parameter is used by using that parameter. Based on the threshold value set for the image, the threshold values for other images are automatically set by the apparatus described in FIG.

As an example of the apparatus, a high-speed image processing analysis apparatus can be used. A first embodiment of the present invention will be described below with reference to the drawings. Figure 2 shows the first and second images,
A graph in which the image area of the portion on the horizontal axis and the density (gray level) are plotted on the vertical axis is shown. FIG. 3 is a graph in which the horizontal axis represents the gray level (brightness) and the vertical axis represents the density distribution of the image, and is a graph of the first image and the second image.

Graph A- of the first image in FIG.
The difference in the most frequent value between the graph B-2 of the second image and the second image,
That is, (Fmax-1)-(Fmax-2) is added to the graph B of the second image in FIG.
It can be seen that binarization can be performed with HR without excess or deficiency.

In FIG. 2, an example in which the density distribution of the second image is high has been described, but conversely, if it is too low, the difference is similarly added to increase the brightness of the image on the monitor screen, that is, By setting the same reference image level,
Can be analyzed. When the difference is not added to the second and subsequent images, the same effect can be obtained by adding the difference to THR.

Next, the operation in the second to fourth embodiments until the image of the measurement sample is binarized by the apparatus shown in FIG. 1 will be described with reference to the flow chart of FIG. When obtaining a target image with an optical microscope, first, a measurement target position of a sample is determined (ST1), and then a measurement region is extracted from the position (ST2) and imaged. The auto stage driver 10 positions the measurement target and determines the area.
After inputting the image of the target area (ST3), shading correction for correcting unevenness of image density due to uneven illumination etc. is performed (ST4), and the density of each pixel forming the image is calculated to express the density distribution frequency. A density histogram is created (ST5). Next, an arbitrary image is taken out as a reference image, and the analyst determines the threshold value of this image by referring to the histogram (ST6). The density of the peak of the histogram representing the characteristics of this image is obtained (ST7). As the density of the peak, the density that is the mode of the density, the density of the peak of the size next to the mode, and the density of the peak that represents the ground (background) are used. Next, the difference between the densities of predetermined peaks is calculated for the histograms of the arbitrary image and the other image, and the value obtained by subtracting this difference from the threshold of the arbitrary image is used as the threshold of the other image (ST8). . Binarization of an arbitrary image and other images is performed by the threshold value thus determined (S
T9).

Next, a second embodiment will be described. The image of interest was taken from the same series of samples or the same sample,
The plurality of images are similar to each other. The analyst sets a threshold value for the first image (arbitrary image) of these, and based on this, the threshold values of the other images are set by the histograms of both images. The same applies to the subsequent examples.

FIG. 5A shows an arbitrary image, and FIG. 5B shows this binary image. Both have the same image.
In the image of (A), stains are present as black spots on a white background paper, and this is a binary image as in (B), and the stain area is measured.

FIG. 6 shows the histogram of the second embodiment,
FIG. 6A is a histogram of an arbitrary image shown in FIG. 5A, and the threshold value T is a value set by the analyst. Figure 6
In (B), regarding the density of the mode value representing the feature in the histogram of another image, the difference obtained by subtracting the density of the mode value of another image from the density of the mode value of any image is represented by ΔX. . By subtracting ΔX from the threshold value T set for an arbitrary image, the distance (density difference) between the mode density and the threshold value in both images can be made the same. As a result, the threshold value T1 of the other image is a value obtained by subtracting ΔX from the threshold value T of the arbitrary image. Although FIG. 6 shows the case where ΔX is positive, subtraction is performed even when it is negative. That is, in the case of a negative value, the absolute value of ΔX is added. Subtraction in the following embodiments has the same meaning.

Next, the second embodiment will be described using another image. FIG. 7 shows an image having a different property from FIG. 5 and its binary image. FIG. 7 (A) is an arbitrary image, which is the case where the gray-colored paper has stains, which occurs in the case of recycled paper or the like.
FIG. 7B is a binary image representing the stain of FIG. 7A.
FIG. 8 shows a case where an image (FIG. 7) different from the image (FIG. 5) shown in the second embodiment is used, and FIG. 8 (A) is the histogram of FIG. 7 (A). FIG. 8 (B) shows a histogram of another image similar to FIG. 7 (A). The density of the most frequent value of both histograms is larger by ΔX in an arbitrary image, and a value obtained by subtracting this ΔX from the threshold value T of the arbitrary image is set as the threshold value T1 of another image. As a result, the distance between the mode density and the threshold value of both images becomes the same.

Next, a third embodiment will be described. The third embodiment uses the peak next to the mode value as the peak of the histogram showing the characteristics of the image. Figure 9
FIG. 9A shows an arbitrary image used in the third embodiment, and FIG.
(B) shows this binary image. The third embodiment is used when there is a gray area on a white background and black points in the gray area are extracted to calculate the area thereof.

FIG. 10 shows the histogram of the third embodiment, and (A) shows the histogram of an arbitrary image shown in FIG. 9 (A). The histogram has three peaks, and is composed of a mode peak having a whiteness of the ground, a second peak having a gray area, and a third peak having a black dot. The threshold value T is set by the analyst in the middle of the densities of the peaks of the gray and black dots in order to obtain a binary image in which black dots are identified from the gray area.

FIG. 10 shows a histogram of the third embodiment. FIG. 10A is an arbitrary image histogram.
FIG. 10B is a histogram of another image similar to an arbitrary image. The density of the second largest peak in both histograms is larger by ΔX in an arbitrary image, and the value obtained by subtracting this ΔX from the threshold value T in the arbitrary image is the threshold value T in other images.
Set to 1. As a result, the distance between the density of the second peak in the histograms of both images and the threshold becomes the same.

Next, a fourth embodiment will be described. In the fourth embodiment, the peak representing the density of the ground is used as the peak of the histogram representing the characteristics of the image. In the second and third examples, the peak having the most frequent value occurred at the density of the ground. However, the density of the ground may not be the most frequent value, and in this case, the peak generated in the density of the ground is set as the characteristic peak.
This is an example. FIG. 11 shows an image of a halftone dot used for a transmission photograph or the like. FIG. 11 (A) shows a halftone dot image in which the peak value of the ground density is the mode value, and FIG. 11 (B) shows a halftone dot image in which the ground peak value is not the most frequent value. In the fourth embodiment, a method of setting a threshold value by using an arbitrary image as the image of FIG. 11A and another image as the image of FIG. 11B will be described.

FIG. 12 shows a histogram of the fourth embodiment. FIG. 12A shows the histogram of the image of FIG. 11A showing an arbitrary image, and FIG. 12B shows the histogram of FIG. 11B showing another image. As shown in FIG. 11 (A), the histogram of the arbitrary image in FIG. 12 (A) has a large peak value of white density representing the background and a small peak value of black density representing the halftone dots. In the histograms of the other images in FIG. 12B, the peak value of the white density representing the background is small, and the peak value of the black density representing the halftone dots is large. Figure 12
A threshold value T for binarizing a halftone dot is set in (A) while looking at FIG. 11 (A) and FIG. 12 (A). When the apparatus sets the threshold value T1 of another image, a range of densities representing white of the background is determined in advance, and the density representing the largest peak generated in this density range is set as the background density. The density of the peak representing the ground in both histograms is ΔX larger in any image,
A value obtained by subtracting this ΔX from a threshold value T of an arbitrary image is set as a threshold value T1 of another image. As a result, the distance between the background density and the threshold in the histograms of both images becomes the same.

[0038]

As described above, the present invention is Analysis can be performed without requiring skill in setting THRs of multiple images.
2. Storing the standard image and THR in the external memory facilitates standardization of work. 3. It is possible to analyze images by eliminating individual differences, image differences, and time differences. 4.
Even if the amount of illumination light varies, the error due to the variation can be eliminated by keeping Fmax constant. 5. Unattended measurement is also possible. And so on.

Further, according to the present invention, when a threshold value is set for an arbitrary image of similar images, the distance between the threshold value and the peak density of the histogram representing the characteristics of the arbitrary image is obtained, and this distance is the same. The threshold of the other image is determined with respect to the density of the peak representing the characteristic of the other image. Therefore, the difference in the densities of the histograms representing the characteristics of both images is obtained, and the value obtained by subtracting this difference from the threshold of any image is used as the threshold of the other image. This produces the same effect as determining the threshold values of other images under substantially the same conditions as determining the threshold value of an arbitrary image. As the peak of the histogram showing the features of both images, the peak showing the mode of the density, the peak of the size next to the mode, or the peak in the density showing the ground is used to match the features of the image. It is possible to obtain a threshold value of another image that has been processed.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an apparatus that realizes a present exemplary embodiment.

FIG. 2 is a graph for explaining the first embodiment, in which the horizontal axis represents the image area of the first image and the second image, and the vertical axis represents the density (gray level).

FIG. 3 is a graph showing a gray level (brightness) on the horizontal axis and an image density distribution on the vertical axis for explaining the first embodiment, and is a graph of a first image and a second image. . It should be noted that reference numerals in FIGS. 2 and 3, A is a graph showing the first image, B is a graph showing the second image, and THR is a threshold value.

FIG. 4 is a flow chart of setting a threshold of another image based on a threshold of an arbitrary image among similar images.

FIG. 5 is a diagram showing an arbitrary image and its binary image used in the second embodiment.

FIG. 6 is a diagram showing a density histogram of the second embodiment.

FIG. 7 is another arbitrary image for explaining the second embodiment and its part 2.
It is a figure which shows a value image.

FIG. 8 is a diagram showing a density histogram of the second embodiment using FIG.

FIG. 9 is a diagram showing an arbitrary image and its binary image for explaining the third embodiment.

FIG. 10 is a diagram showing a density histogram according to the third embodiment.

FIG. 11 is a diagram showing an arbitrary image and another image for explaining a fourth embodiment.

FIG. 12 is a diagram showing a density histogram of a fourth embodiment.

Claims (6)

[Claims] 1. A method of displaying a digital image of a printed matter or the like including an image to be analyzed on a monitor screen, binarizing the digital image by a threshold value according to shading, and performing image analysis, in an arbitrary first image. The threshold value is determined based on the reference portion and the mode (Fmax-1) of the density distribution of the image is determined.
And the difference between the mode (Fmax−1) and the mode (Fmax−N) of the density distribution of the second and subsequent images,
In the binarization of the second and subsequent images, after adding to each pixel of the second and subsequent images, the image is analyzed based on the threshold value. 2. A method for displaying a digital image of a printed matter or the like including an image to be analyzed on a monitor screen, binarizing the digital image with a threshold value according to shading, and performing image analysis, in an arbitrary first image. The threshold value is determined based on the reference portion and the mode (Fmax-1) of the density distribution of the image is determined.
And the difference between the mode (Fmax−1) and the mode (Fmax−N) of the density distribution of the second and subsequent images,
An image analysis method, wherein in the binarization of the second and subsequent images, analysis is performed based on the threshold value added to the threshold value. 3. An image analysis method for creating a histogram of a grayscale distribution area according to preset grayscale levels of a plurality of similar grayscale images, and determining a threshold value of a corresponding image based on the histogram. The threshold value of the image is determined, and the difference between the density of the predetermined peak of the histogram representing the feature of the arbitrary image and the density of the predetermined peak of the histogram representing the feature of another image is calculated to obtain the difference of the arbitrary image. An image analysis method, wherein a density obtained by subtracting the difference from a threshold value is used as a threshold value of another image. 4. The image analysis method according to claim 3, wherein the density of the predetermined peak is a density representing the mode of the density distribution. 5. The image analysis method according to claim 3, wherein the density of the predetermined peak is a density representing the second largest peak value after the mode of the density distribution. 6. The image analysis method according to claim 3, wherein the density of the predetermined peak is the density of a peak near the density representing the background of the image.