JPH0683959A - Area division system for image - Google Patents

Abstract

PURPOSE:To perform an image dividing process wherein the influence of a noise is small by setting a threshold value for pixels regarding the edge part of the area division. CONSTITUTION:The image area division system which divides an input image 1 into plural areas corresponding to density values is equipped with an edge detecting means 2, a minimum value filter 4, a cumulative histogram means 3 which calculates the cumulative histogram of the input image 1 on the edge, and a cumulative histogram means 5 which calculates the cumulative histogram of the image after a minimum value filtering process on the edge. Further, the system is equipped with a histogram difference calculating means 6 which calculates the difference between those two kind of cumulative histograms, density by density, a multiple threshold value calculating means 7 which obtains plural density values corresponding to the peaks of the obtained histogram, an area dividing means 8 which divides the image into areas by using the density values as threshold values, and a labeling means L which labels connection components in the respective areas.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an area dividing system for dividing an image into areas based on the density value and the edge position.

[0002]

2. Description of the Related Art For example, as shown in FIG. 26A, when there is an image having a light density area a1, an intermediate density area a2, and a dark density area a3, this image is used as a density value and edge. When dividing the image into each area based on the position of
Conventionally, the method as shown in FIG. 27 (A) or FIG. 27 (B) has been used.

In the method of FIG. 27A, first, the magnitude of the differential is calculated in each pixel of the input image, and then the pixel having the differential value larger than a predetermined value is detected as an edge between regions. Next, a histogram of the density values of the input image is calculated only for the pixels on the edge. In this histogram, each density value corresponds to the number of pixels on the edge having that density.

Next, by selecting a plurality of density values corresponding to the peaks of this histogram, multiple threshold values g1, g2, ...
gN. And finally, in the input image, the density is 0
The regions are divided into regions such that g1, g1 to g2, ..., GN to gmax, and the connected components are labeled in the respective regions to perform region division. Here, gmax is the maximum value that the density can take.

Further, in the method of FIG. 27B, after calculating the magnitude of the differential in each pixel of the input image, the differential value larger than a predetermined value is detected as an edge, and then the edge is detected. A two-dimensional (2D) histogram relating to the density value of the input image and the magnitude of the differential is calculated for only the pixels.

In this histogram, the number of pixels on the edge having the density and the magnitude of the differential corresponds to each density value and the magnitude of the differential. Next, a plurality of peaks in this histogram are selected, and the density values corresponding to the peaks are set as the multiple threshold values g1, g2 ... gN. The subsequent processing is the same as in the case of FIG.

[0007]

FIG. 26 is a view for explaining the problems of the conventional method. FIG. 26A shows an input image, and FIG. 26B shows a cross section of the density value along the straight line 1 in FIG. The thick lines in the figure are edges. The input image is as shown in FIG. a1, a2,
It is assumed to consist of three areas a3. In addition, as shown in FIG. 26 (B), the respective regions are f1, f2,
It has an average density of f3, and the density value of each pixel flickers randomly around the average value due to noise or the like.

In the method shown in FIG. 27 (A),
At the edge between the regions a1 and a2, f1 + d1 or f2 + d
The density histogram on the edge is f1 and f2 because the density is 2 (however, d1 and d2 flickers randomly).
Has a peak.

However, using either f1 or f2 as a threshold,
Even if the area shown in FIG. 26A is divided, an appropriate result cannot be obtained due to the influence of noise. In the method shown in FIG. 27B as well, the density values corresponding to the peak values of the two-dimensional histogram on the edges are f1 and f2, and an appropriate result cannot be obtained due to the influence of noise. There are some difficulties.

As described above, the conventional method has a problem in that, when the density changes sharply at the boundary between the areas, the density threshold for dividing the area cannot be properly determined. In addition,
When the density changes relatively slowly, the density at the edge is not only f1 + d1 or f2 + d2 but also f1 and f.
Since the number of cases having an intermediate density of 2 is increased, the above-mentioned difficulty is alleviated, but the application range in which the image cannot be used unless it is an image that changes gently is not preferable.

SUMMARY OF THE INVENTION It is therefore an object of the present invention to solve the above-mentioned problems and to provide a region segmentation method for an image in which a threshold value is set between f1 and f2 and between f2 and f3.

[0012]

Therefore, in the present invention,
When there are areas a1, a2, and a3 as shown in FIG. 2A, it is assumed that the areas are divided using the density value f shown in FIG. 2B as a threshold value. That is, it is assumed that the region is divided into a region having a density lower than f and a region having a density higher than f. Now consider the probability that the boundaries of these regions are on edges. When f has a value just between f1 and f2, there is a high probability that the boundary exists on the edge, but when f is close to f1 or f2, the boundary is above the edge because it is easily affected by noise. The probability of riding on is small.

Therefore, if the threshold f is set so that the probability that the boundary exists on the edge becomes large, the regions can be divided more appropriately. Therefore, in the present invention, a plurality of thresholds are calculated so that the probability that the boundary is on the edge is large, the regions are divided by these thresholds, and then the regions are divided by labeling the connected components.

A basic embodiment of the present invention will be described with reference to FIG. When dividing an area by the existing threshold value f, the number of boundary points is the same as the cumulative histogram of the input image (corresponding to the total sum of pixels having a density larger than the density value for each density) and the minimum value filter processing. Difference of the cumulative histogram of the image, or the difference of the cumulative histogram of the image after the maximum value filtering and the input image, or the difference of the cumulative histogram of the image after the maximum value filtering and the image after the minimum value filtering Obtained by dividing by 2 (JP-A-1-162992, JP-A1-1).
(See Japanese Patent No. 84583).

As shown in FIG. 1, an edge detection unit 2 detects an edge of an input image 1 input and held by an image input unit such as a VT camera. That is, after the differentiation, those having a differential value equal to or larger than the threshold value are extracted to detect the edge. Then, the cumulative histogram unit 3 creates a cumulative histogram for the input image on this edge.

Further, the input image 1 is subjected to the minimum value filter processing in the minimum value filter unit 4. Of these, the cumulative histogram of only the edge portion is created by the cumulative histogram unit 5. Then, the difference detection unit 6 of the histogram finds the difference between the cumulative histogram of the edge part with respect to the input image created by the cumulative histogram unit 3 and the cumulative histogram of the edge part subjected to the minimum value filtering created by the cumulative histogram unit 5. . This leaves multiple peaks.

The multiple threshold value calculating section 7 detects the density values g1, g2, ... gN corresponding to the plurality of peaks of the histogram obtained by the difference calculating section 6 of the histogram, and sets them as a plurality of threshold values. . Then, in the area dividing unit 8,
By processing the input image 1 using these density values as threshold values, the input image is processed to have the densities of 0 to g1, g1 to g2, ...
It is possible to divide the input image into regions by dividing the input image into regions of up to gmax and then labeling the connected components by the labeling unit L.

[0018]

In this way, since the threshold value can be set only for the pixels on the edge, it is possible to perform accurate area division with less influence of noise.

[0019]

EXAMPLES Examples of the present invention will be described in detail below. FIG. 1 is also an embodiment of the present invention, but since its configuration has been described in detail above, it uses the maximum value filter described above with reference to FIGS. 3 and 4, and uses the maximum value filter and the minimum value filter. Will be described.

In FIG. 3, a maximum value filter 9 is used. The input image 1 is subjected to the same processing as that shown in FIG. 1 in the edge detection unit 2 and the cumulative histogram unit 3 to create a cumulative histogram for the input image on the edge.

The input image is subjected to maximum value filtering in the maximum value filter section 9. Of these, the cumulative histogram of only the edge portion is created by the cumulative histogram unit 5. Then, the difference calculation unit 6 of the histogram obtains the difference between the cumulative histogram of the edge part with respect to the input image created by the cumulative histogram unit 3 and the cumulative histogram of the maximum-value filtered edge part created by the cumulative histogram unit 5. Based on the obtained plurality of peak values, the region division is performed as in the case of FIG. 1, the connected component is labeled by the labeling unit L, and the input image is divided into regions.

FIG. 4 shows an example in which the minimum value filter 4 and the maximum value filter 9 are used. The edge of the input image is detected by the edge detector 2 as in the case of FIGS. 1 and 3. Further, the input image 1 is subjected to minimum value filter processing and maximum value filter processing by the minimum value filter 4 and the maximum value filter 9, respectively.

Cumulative histograms of the edge portions subjected to the minimum value filter processing and the maximum value processing are created by the cumulative histogram unit 3 and the cumulative histogram unit 5, respectively.
The histogram difference calculation unit 6 obtains these differences, and the multiple threshold calculation unit 7 and the region division unit 8 perform the same processing based on the density values g1, g2, ... Corresponding to the obtained plurality of peaks. In the histogram difference calculation unit 6, even if the difference is halved, the position of the peak does not change, so this division can be omitted.

Next, the second type of invention of the present invention will be described with reference to FIGS.
FIG. 7 will be described. The second type invention uses the histogram representing the number of boundary points on an edge in the first type invention divided by a histogram representing the number of boundaries in the entire image.

Since the number of boundary points in the entire image becomes large due to the influence of noise when the density is close to f1 or f2 in FIG. 2, division by this number makes the peak of the histogram steep and more stable. It becomes possible to detect peaks.

The number of boundary points of the entire image is the difference between the cumulative histograms of the input image and the image after the minimum value filtering, and the cumulative histograms of the image after the maximum value filtering and the input image, respectively, for the entire image. And the difference between the cumulative histograms of the image after the maximum value filtering and the image after the minimum value filtering are divided by two.

FIG. 5 shows an embodiment corresponding to the above. The edge of the input image 1 is detected by the edge detection unit 2 after the differentiation, and the differential value of which is greater than or equal to a threshold value is extracted. Further, the input image 1 is subjected to the minimum value filter processing by the minimum value filter 4.

The cumulative histogram unit 3 creates a cumulative histogram of the edge-detected part of the input image 1, and the cumulative histogram unit 5 creates a cumulative histogram of the edge-detected part of the input image subjected to the minimum value filtering.
The cumulative histogram unit 10 creates a cumulative histogram of the input image 1, and the cumulative histogram unit 11 creates a cumulative histogram for the minimum-value filtered input image.

Then, the difference between the cumulative histogram of the edge part of the input image 1 created by the cumulative histogram part 3 and the cumulative histogram of the edge part processed by the minimum value filter created by the cumulative histogram part 5 is calculated as a difference between the histograms. Obtained in part 12. Further, the difference calculation unit 13 calculates the difference between the cumulative histogram of the input image created by the cumulative histogram unit 10 and the cumulative histogram of the input image created by the cumulative histogram unit 11 subjected to the minimum value filtering.

In the histogram quotient calculation unit 14, the data relating to the cumulative histogram of the edge portion obtained from the difference calculation unit 12 is quotient with the data relating to the number of boundary points of the entire image obtained from the difference calculation unit 13, The multiple threshold value calculating unit 7 detects the density values g1, g2, ...

Then, the area dividing section 8 processes the input image using these density values as threshold values and sets the density values from 0 to 0.
The input image can be divided into regions by dividing the regions into g1, g1 to g2 ... And labeling the connected components.

An embodiment corresponding to the above will be described with reference to FIG. The edge of the input image 1 is detected by the edge detection unit 2 as described above, and the maximum value filter 9 performs maximum value filtering.

The cumulative histogram unit 3 creates a cumulative histogram of the edge-detected part of the input image 1, and the cumulative histogram unit 5 creates a cumulative histogram of the edge-detected part of the maximum-value processed input image.

Then, the difference between the cumulative histogram of the edge part of the input image 1 created by the cumulative histogram part 3 and the cumulative histogram of the edge part subjected to the maximum value filtering created by the cumulative histogram part 5 is calculated as a difference between the histograms. Obtained in part 12. Further, the difference calculation unit 13 calculates the difference between the cumulative histogram of the input image created by the cumulative histogram unit 10 and the cumulative histogram of the input image created by the cumulative histogram unit 11 subjected to the maximum value filter process.

In the histogram quotient calculation unit 14, the data relating to the cumulative histogram of the edge portion obtained from the difference calculation unit 12 is quoted by the data of the number of boundary points of the entire image obtained from the difference calculation unit 13 to obtain the data. .. corresponding to the plurality of peaks are detected by the multiple threshold value calculation unit 7, and the input image is divided into regions as described above.

An embodiment corresponding to the above will be described with reference to FIG. The edge of the input image 1 is detected by the edge detecting unit 2 as described above, the minimum value filter 4 performs the minimum value filter processing, and the maximum value filter 9 performs the maximum value processing.

The cumulative histogram unit 3 creates a cumulative histogram of the edge portion of the input image 1 that has been subjected to the minimum value filtering, and the cumulative histogram unit 5 creates a cumulative histogram of the edge portion of the input image that has undergone the maximum value filtering. Then, the cumulative histogram unit 10 creates a cumulative histogram of the input image subjected to the minimum value filtering, and the cumulative histogram unit 11 creates a cumulative histogram of the input image subjected to the maximum value filtering.

Then, the difference between the cumulative histogram of the edge-filtered edge portion of the input image 1 created by the cumulative histogram unit 3 and the cumulative histogram of the maximum-filtered edge portion created by the cumulative histogram unit 5 is calculated. The difference is calculated by the histogram difference calculation unit 12. Further, the difference between the cumulative histogram of the minimum-value filtered input image created by the cumulative histogram unit 10 and the cumulative histogram of the maximum-value filtered input image created by the cumulative histogram unit 11 is calculated as a histogram difference calculation unit. 1
Find in 3.

Then, in the histogram quotient calculation unit 14, the data relating to the cumulative histogram of the edge portion obtained from the difference calculation unit 12 is quoted by the data of the number of boundary points of the entire image obtained from the difference calculation unit 13 and obtained. .. corresponding to the plurality of peaks are detected by the multiple threshold value calculation unit 7, and the input image is divided into regions as described above.

Next, the third type invention of the present invention will be described with reference to FIGS.
13 will be described. Since edge detection is performed in the first or second type invention, the reliability of area division depends on the quality of edge detection. The edge is detected from the input image as a point having a magnitude of differentiation larger than a predetermined threshold value in each pixel. If this threshold is not appropriate, the accuracy of edge detection will deteriorate.

In order to solve this problem, the histogram is calculated with respect to an arbitrary threshold for edge detection. When edge detection is performed with a certain threshold, the number of boundary points on the edge is the two-dimensional cumulative histogram of the magnitude of the differential and the density of the input image, and the two-dimensional cumulative histogram of the magnitude of the differential and the density after the minimum value filtering. Can be expressed by the difference (see JP-A-1-239678). Of course, the maximum value filter may be used instead of the minimum value filter, or both the minimum value filter and the maximum value filter may be used. Then, as shown in FIGS. 11 to 13, the peak can be made sharper by dividing the histogram by the number of boundary points in the entire image.

An embodiment of the third type will be described with reference to FIG. The input image 1 is differentiated and the magnitude of the differential is detected by the differential magnitude detection unit 20. Then, the 2D cumulative histogram unit 21 creates a two-dimensional cumulative histogram for the differential value and the density of the input image corresponding thereto.

The input image 1 is subjected to the minimum value filter processing by the minimum value filter 4. Then, the 2D cumulative histogram unit 22 creates a two-dimensional cumulative histogram based on the input image subjected to the minimum value filter processing and the magnitude of the differential.

The difference between the two two-dimensional cumulative histograms thus obtained is obtained by the 2D histogram difference calculating section 23, and the densities g1 and g2 corresponding to the obtained plurality of peaks are obtained.
... are detected by the multiple threshold value calculating unit 7 and the input image is divided by the region dividing unit 8 as described above.

FIG. 9 shows an embodiment in which the maximum value filter is used instead of the minimum value filter in FIG. The input image 1 is differentiated and the magnitude of the differential is detected by the differential magnitude detection unit 20. Then, the 2D cumulative histogram unit 21 creates a two-dimensional cumulative histogram for the differential value and the density of the input image corresponding to the differential value.

Further, the input image 1 is subjected to the maximum value filter processing by the maximum value filter 9. Then, the two-dimensional cumulative histogram based on the density of the input image subjected to the maximum value filtering and the magnitude of the differential is the 2D cumulative histogram unit 2
Created by 2.

The difference between the two two-dimensional cumulative histograms thus obtained is calculated by the 2D histogram difference calculating section 23, and the densities g1 and g corresponding to the plurality of obtained peaks are calculated.
.. is detected by the multiple threshold calculation unit 7, and the input image is divided by the region division unit 8 in the same manner as described above.

FIG. 10 shows an embodiment using a minimum value filter and a maximum value filter. The input image 1 is differentiated and the magnitude of the differential is detected by the differential magnitude detection unit 20. Then, by the 2D cumulative histogram unit 21,
A two-dimensional cumulative histogram for the density of the input image that has been subjected to the minimum value filtering by the differential value and the minimum value filter 4 is created.

The input image is subjected to the maximum value filter processing by the maximum value filter 9, and the two-dimensional cumulative histogram based on the density of the processed input image and the magnitude of the differential is 2.
It is created by the D cumulative histogram unit 21.

The difference between the two two-dimensional cumulative histograms thus obtained is calculated by the 2D histogram difference calculating section 23, and the densities g1 and g corresponding to the obtained plurality of peaks are calculated.
.. is detected by the multiple threshold value calculating unit 7, and the input image is divided by the region dividing unit 8 in the same manner as described above.

In FIG. 11, the magnitude of the differential of the input image 1 is detected by the differential magnitude detecting section 20 in the same manner as described above, and the 2D cumulative histogram section 21 detects the differential value and the corresponding density of the input image. A dimension cumulative histogram is created. Further, the input image 1 is subjected to the minimum value filter processing by the minimum value filter 4.

The 2D cumulative histogram unit 22 creates a two-dimensional cumulative histogram of the differential value and the density of the image subjected to the minimum value filter processing. The cumulative histogram unit 24 creates a cumulative histogram of the density of the input image 1, and the cumulative histogram unit 25 creates a cumulative histogram of the density of the image output by the minimum value filter processing.

Then, the differential value created by the 2D cumulative histogram unit 21 and the two-dimensional cumulative histogram for the input image, the differential value created by the 2D cumulative histogram unit 22 and the density of the image output by the minimum value filtering process The difference between the two-dimensional cumulative histograms is obtained by the 2D histogram difference calculation unit 23. This gives the number of boundary points on the edge portion.

Further, the difference between the cumulative histogram of the densities of the input image 1 created by the cumulative histogram unit 24 and the cumulative histogram of the densities of the images subjected to the minimum value filtering created by the cumulative histogram unit 25 is calculated as a difference between the histograms. Obtained by the unit 26, the number of boundary points in the entire image is obtained.

In the quotient calculating unit 27 of the 2D histogram, the output of the difference calculating unit 23 of the 2D histogram, that is, the number of boundary points on the edge, is calculated by the difference calculating unit 2 of the histogram.
6 output, that is, by dividing by the number of boundary points in the entire image, the densities g1, g2 ...
To get As a result, the input image can be divided into regions in the multiple threshold calculation unit 7 and the region division unit 8 as described above.

In the embodiment shown in FIG. 12, the maximum value filter 9 is used instead of the minimum value filter 4 in FIG. 11, and since it corresponds to FIG. 9, detailed description thereof will be omitted.

The embodiment shown in FIG. 13 uses the minimum value filter 4 and the maximum value filter 9. The differential magnitude detection unit 20 detects the differential magnitude of the input image 1, and the 2D cumulative histogram unit 21 creates a two-dimensional cumulative histogram of the differential value and the density of the image subjected to the minimum value filtering corresponding to the differential value. To be done. The 2D cumulative histogram unit 22 creates a two-dimensional cumulative histogram of the differential value and the image density subjected to the maximum value filter processing corresponding to the differential value.

The cumulative histogram unit 24 creates a cumulative histogram of the density of the image subjected to the minimum value filter processing. Then, the cumulative histogram unit 25 creates a cumulative histogram of the density of the image subjected to the maximum value filter processing.

The difference between the two-dimensional cumulative histogram of the differential value created by the 2D cumulative histogram section 21 and the density of the image subjected to the minimum value filtering is calculated by the 2D histogram difference calculation section 23. The number of boundary points of is obtained.

Further, the difference between the cumulative histogram of the densities of the image processed by the minimum value filter created by the cumulative histogram unit 24 and the cumulative histogram of the densities of the image processed by the maximum value filter created by the cumulative histogram unit 25 is calculated. The histogram difference calculation unit 26 obtains the number of boundary points in the entire image.

Then, in the 2D histogram quotient calculating unit 27, the output of the 2D histogram difference calculating unit 23 is divided by the output of the histogram difference calculating unit 26, so that the densities g1, g2 ... To get As a result, the input image can be divided into regions in the multiple threshold calculation unit 7 and the region division unit 8 as described above.

The fourth type invention of the present invention will be described with reference to FIGS. In these, the histogram is not divided by the number of boundary points in the entire image, but by the sum of the number of boundary points and the number of edges in the entire image.
That is, if the threshold value for detecting edges is made smaller, the number of edges becomes larger as the threshold value becomes smaller, but noise becomes easier to pick up as edges instead. Therefore, when the number of edges becomes large, it is preferable to include and divide the number of edges so that the threshold value for area division decreases accordingly. With this, when the threshold for edge detection is too small, it is possible to avoid the case where a large number of unessential edges are erroneously detected due to the influence of noise. A weighted linear sum may be used instead of the simple sum of the number of boundary points and the number of edges in the entire image.

An embodiment of the fourth type will be described with reference to FIG. In FIG. 14, in the embodiment of FIG. 11, the cumulative histogram unit 28 is provided and the differential magnitude detecting unit 2 is provided.
A cumulative histogram of differential values output from 0 is created. This makes it possible to obtain the number of edges in the entire image. Therefore, the quotient calculation unit 2 of the 2D histogram
9, the number of boundary points on the edges output from the 2D histogram difference calculation unit 23 is defined as the number of boundary points in the entire image output from the histogram difference calculation unit 26 and the number of edges in the entire image. It is possible to avoid a case in which a large number of non-essential edges are erroneously detected due to the influence of noise.

Then, the quotient calculation unit 29 of the 2D histogram
With the densities g1, g2, ... Obtained as described above, the input image can be divided into regions by the multiple threshold calculation unit 7 and the region division unit 8 as described above.

The embodiment shown in FIG. 15 is different from the embodiment shown in FIG. 12 in that a cumulative histogram unit 28 is provided to obtain the number of edges in the entire image, and a quotient calculation unit 29 of the 2D histogram calculates the difference between the 2D histograms. The number of boundary points on the edges output from the calculation unit 23 is calculated as the number of boundary points in the entire image output from the histogram difference detection unit 26 and the number of edges in the entire image output from the cumulative histogram unit 28. , And the other configurations are the same as those in FIG. 12, and the description thereof will be omitted.

In the embodiment shown in FIG. 16, in the embodiment shown in the thirteenth embodiment, a cumulative histogram unit 28 is provided to obtain the number of edges in the entire image, and a quotient calculation unit 29 of the 2D histogram calculates a difference calculation unit of the 2D histogram. The number of boundary points on the edge output from 23 is the sum of the number of boundary points in the entire image output from the histogram difference detection unit 26 and the number of edges in the entire image obtained from the cumulative histogram unit 28. The configuration is the same as that shown in FIG. 13, and the description thereof is omitted.

In the third type and the fourth type, a plurality of peak values of the obtained two-dimensional histogram are detected, the regions are divided by the density values corresponding to those peaks, and then the labels of the connected components are obtained. It can be divided into areas by attaching.

When finding the peak of the two-dimensional histogram, in order to obtain a more stable one with respect to the change in the magnitude of the differential, it is necessary to find a ridge parallel to the axis of the differential magnitude in the two-dimensional histogram. Is preferred.

In the third type or the fourth type, the cumulative histogram relating to the density of the input image, the density of the image after the minimum value filtering, or the density of the image after the maximum value filtering is performed. Is zero in the two-dimensional cumulative histogram of the magnitude of the differential and the input image, or the magnitude of the differential and the image after the minimum value filtering, or the density of the differential and the image after the maximum value filtering. , The density of the input image, or the density of the image after the minimum value filtering, or the cumulative histogram of the density of the image after the maximum value filtering, instead of the differential magnitude and the density of the input image. , Or the density of the image after differential magnitude and minimum value filtering, or the image after differential magnitude and maximum value filtering In the two-dimensional accumulated histogram of density may be used histogram portion size of the derivative is zero. As a result, the amount of calculation can be reduced.

In the fourth type, the cumulative histogram concerning the magnitude of the differential is equal to the partial histogram in which the density is zero in the two-dimensional cumulative histogram relating to the magnitude of the differential and the density of the input image. Instead of the cumulative histogram concerning the magnitude, a partial histogram in which the density is zero in the two-dimensional histogram concerning the magnitude of the differential and the density of the input image may be used. Also in this case, the calculation amount can be limited.

FIG. 17 shows a histogram obtained by using the histogram of the differential magnitude and the portion of the input image where the differential magnitude is zero in FIG. In FIG. 17, the differential magnitude and the portion of the input image where the differential magnitude is zero,
The difference calculation unit 30 of the partial histogram obtains the difference in the two-dimensional cumulative histogram between the differential magnitude and the portion where the differential magnitude is zero for the image subjected to the minimum value filtering, and the same area division processing is performed.

FIG. 18 shows a two-dimensional cumulative histogram of the differential magnitude and the portion of the input image having the differential magnitude of zero in FIG. 12, and the differential magnitude and the maximum magnitude filtered image having the differential magnitude of zero. The difference calculation unit 30 of the partial histogram obtains the difference between the above areas, and the same area division processing is performed.

FIG. 19 shows the two-dimensional cumulative histogram of the differential magnitude and minimum value filtered image in FIG. 13 in which the differential magnitude is zero, and the differential magnitude and maximum value filtered image. The difference between the two-dimensional cumulative histograms of the portion where the magnitude of the differential is zero is calculated by the difference calculation unit 30 of the partial histograms, and the same area division processing is performed.

FIG. 20 shows a two-dimensional cumulative histogram of the differential magnitude and the portion of the input image where the differential magnitude is zero in FIG. 14, and the differential magnitude and the minimum magnitude filtered image having the differential magnitude of zero. The difference from the two-dimensional cumulative histogram of the part is calculated by the difference calculation unit 30 of the partial histogram, and the same area division processing is performed.

FIG. 21 is a two-dimensional cumulative histogram of the differential magnitude and the portion of the input image where the differential magnitude is zero in FIG. 15, and the differential magnitude and the maximum magnitude filtered image have a differential magnitude of zero. The difference from the two-dimensional cumulative histogram of the part is calculated by the difference calculation unit 30 of the partial histogram, and the same area division processing is performed.

FIG. 22 shows the two-dimensional cumulative histogram of the portion of the differential magnitude and the minimum value filtered in FIG. 16 where the differential magnitude is zero, and the differential magnitude and the maximum filtered image. The difference calculation unit 30 of the partial histogram finds the difference between the two-dimensional cumulative histogram in which the magnitude of the differential is zero, and the same area division processing is performed.

23 to 25 are respectively FIGS. 14 to 1.
6, the cumulative histogram unit 28 related to the magnitude of the differential is not used, and a partial histogram in which the density is zero in the two-dimensional cumulative histogram related to the magnitude of the differential and the density of the input image is used to perform the same area division processing. It is something to do.

[0078]

According to the present invention, when the input image is divided into areas for density, the threshold value can be set only for the pixels on the edges of the input image, so that accurate area division with less influence of noise can be performed. It can be carried out.

[Brief description of drawings]

FIG. 1 is a basic embodiment of the present invention.

FIG. 2 is a diagram illustrating the basic concept of the present invention.

FIG. 3 is a second embodiment of the present invention.

FIG. 4 is a third embodiment of the present invention.

FIG. 5 is a fourth embodiment of the present invention.

FIG. 6 is a fifth embodiment of the present invention.

FIG. 7 is a sixth embodiment of the present invention.

FIG. 8 is a seventh embodiment of the present invention.

FIG. 9 is an eighth embodiment of the present invention.

FIG. 10 is a ninth embodiment of the present invention.

FIG. 11 is a tenth embodiment of the present invention.

FIG. 12 is an eleventh embodiment of the present invention.

FIG. 13 is a twelfth embodiment of the present invention.

FIG. 14 is a thirteenth embodiment of the present invention.

FIG. 15 is a fourteenth embodiment of the present invention.

FIG. 16 is a fifteenth embodiment of the present invention.

FIG. 17 is a sixteenth embodiment of the present invention.

FIG. 18 is a seventeenth embodiment of the present invention.

FIG. 19 is an eighteenth embodiment of the present invention.

FIG. 20 is a nineteenth embodiment of the present invention.

FIG. 21 is a twentieth embodiment of the present invention.

FIG. 22 is a twenty-first embodiment of the present invention.

FIG. 23 is a twenty-second embodiment of the present invention.

FIG. 24 is a twenty third embodiment of the present invention.

FIG. 25 is a twenty-fourth embodiment of the present invention.

FIG. 26 is an explanatory diagram of area division.

FIG. 27 is an explanatory diagram of a conventional example.

[Explanation of symbols]

 1 Input Image 2 Edge Detection Section 3 Cumulative Histogram Section 4 Minimum Value Filter 5 Cumulative Histogram Section 6 Histogram Difference Calculation Section 7 Multiple Threshold Calculation Section 8 Area Division Section L Labeling Section

Claims (19)

[Claims] 1. An area dividing method of an image for dividing an input image into a plurality of areas according to a density value, and an edge detecting unit (2) for detecting an edge of the input image, and a minimum value applying a minimum value filtering process to the input image. A value filter (4), a cumulative histogram means (3) for calculating a cumulative histogram of the input image on the edge, a cumulative histogram means (5) for calculating a cumulative histogram of the image after the minimum value filtering on the edge, A histogram difference calculating means (6) for calculating a difference for each density of these two types of cumulative histograms, and a multiple threshold value calculating means (7) for obtaining a plurality of density values corresponding to the peaks of the histogram obtained by the histogram A region dividing means (8) for dividing the image into regions by using the plurality of density values as thresholds, and a label of connected components in each region. Image Segmentation method characterized by comprising labeling means for performing with (L). 2. In an image area dividing method for dividing an input image into a plurality of areas according to density values, an edge detecting means for detecting an edge of the input image, and a maximum value filter for applying a maximum value filtering process to the input image. , A cumulative histogram means for calculating a cumulative histogram of the input image on the edge, a cumulative histogram means for calculating a cumulative histogram of the image after the maximum value filter processing on the edge, and a difference for each density of these two types of cumulative histograms. Histogram difference calculating means for calculating; multiple threshold value means for obtaining a plurality of density values corresponding to the peaks of the histogram obtained by this; area dividing means for dividing the image into areas using the plurality of density values as threshold values; It is characterized by comprising labeling means for labeling the connected components in each region. Image region division method of the. 3. An image area dividing method for dividing an input image into a plurality of areas according to density values, and an edge detecting unit for detecting an edge of the input image, and a minimum value filter for applying a minimum value filtering process to the input image. , A maximum value filter that applies maximum value filtering to the input image, a cumulative histogram means that calculates the cumulative histogram of the image after the minimum value filtering on the edge, and a cumulative histogram of the image after the maximum value filtering on the edge. A cumulative histogram calculating unit, a histogram difference calculating unit that calculates a density difference between these two types of cumulative histograms, a multiple threshold calculating unit that obtains a plurality of density values corresponding to the peaks of the histogram obtained by these, Region dividing means for dividing an image into regions by using a plurality of density values as thresholds, and each region A region segmentation system for an image, comprising a labeling means for labeling connected components in a region. 4. An area dividing method of an image for dividing an input image into a plurality of areas according to a density value, an edge detecting means for detecting an edge of the input image, and a minimum value filter for applying a minimum value filtering process to the input image. , A cumulative histogram means for calculating a cumulative histogram of the input image on the edge, a cumulative histogram means for calculating a cumulative histogram of the image after the minimum value filter processing on the edge, and a difference for each density of these two types of cumulative histograms. A histogram difference calculating means for calculating, a cumulative histogram means for calculating a cumulative histogram of the input image, a cumulative histogram calculating means for calculating a cumulative histogram of the image after the minimum value filter processing, and a density of each of these two types of cumulative histograms. Difference calculation means for calculating a difference between the two A histogram quotient calculating means for calculating a quotient for each density of the histogram obtained by the histogram difference calculating means, a multiple threshold value calculating means for obtaining a plurality of density values corresponding to the peaks of the histogram obtained thereby, and a plurality of these An area dividing method for an image, comprising: an area dividing unit that divides an image into areas using the density value of 1 as a threshold, and a labeling unit that labels connected components in each area. 5. An area dividing method for dividing an input image into a plurality of areas according to a density value, an edge detecting unit for detecting an edge of the input image, and a maximum value filter for applying a maximum value filtering process to the input image. A cumulative histogram calculating means for calculating the cumulative histogram of the input image on the edge, a cumulative histogram calculating means for calculating the cumulative histogram of the image after the maximum value filter processing on the edge, and Histogram difference calculating means for calculating the difference, cumulative histogram calculating means for calculating the cumulative histogram of the input image, cumulative histogram calculating means for calculating the cumulative histogram of the image after the minimum value filtering, and these two types of cumulative histograms And a histogram difference calculation means for calculating the difference for each concentration of A histogram quotient calculating means for calculating a quotient for each density of the histograms obtained by the difference calculating means of the two histograms; and a multiple threshold value calculating means for obtaining a plurality of density values corresponding to the peaks of the histograms obtained by this. An image region dividing method comprising: a region dividing unit that divides an image into regions using these density values as thresholds; and a labeling unit that labels connected components in each region. 6. An image area dividing method for dividing an input image into a plurality of areas according to a density value, and an edge detecting unit for detecting an edge of the input image, and a minimum value filter for applying a minimum value filtering process to the input image. , A maximum value filter that applies maximum value filtering to the input image, a cumulative histogram means that calculates the cumulative histogram of the image after the minimum value filtering on the edge, and a cumulative histogram of the image after the maximum value filtering on the edge. A cumulative histogram means for calculating, a histogram difference calculating means for calculating a density difference between these two types of cumulative histograms, a cumulative histogram means for calculating a cumulative histogram of the image after the minimum value filtering, and a cumulative histogram means for calculating the minimum value filtering A cumulative histogram calculating means for calculating a cumulative histogram of an image, and A histogram difference calculating means for calculating a difference for each density of the two types of cumulative histograms; and a histogram quotient calculating means for calculating a quotient for each density of the histograms obtained by the difference calculation means for the two histograms. Multiple threshold value calculating means for obtaining a plurality of density values corresponding to the peaks of the histogram, area dividing means for dividing the image into areas by using the plurality of density values as thresholds, and labeling of connected components in each area An image area dividing method, characterized by comprising a labeling means. 7. A region size dividing method of an image for dividing an input image into a plurality of regions according to a density value, and a differential magnitude detecting means for detecting a differential magnitude at each pixel of the input image, and a minimum for the input image. A minimum value filter for performing a value filtering process, a two-dimensional cumulative histogram means for calculating a two-dimensional cumulative histogram relating to the magnitude of the differential and the density of the input image, and a two-dimensional cumulative relating to the magnitude of the differential and the density of the image after the minimum value filtering. A two-dimensional cumulative histogram means for calculating a histogram, a two-dimensional histogram difference calculating means for calculating a difference between density and a differential magnitude of these two types of two-dimensional cumulative histograms, and a peak of the two-dimensional histogram obtained thereby. And a multiple threshold value calculating means for obtaining a plurality of density values corresponding to the Region dividing means and the image area division method, characterized by comprising a labeling means for carrying out labeling of connected components in the respective regions divided into. 8. An area dividing method of an image for dividing an input image into a plurality of areas according to density values, and a differential magnitude detecting means for detecting a differential magnitude at each pixel of the input image, A maximum value filter that applies maximum value filtering, a two-dimensional cumulative histogram means that calculates a two-dimensional cumulative histogram related to the magnitude of the differential and the density of the input image, and a two-dimensional related to the magnitude of the differential and the density of the image after maximum value filtering A two-dimensional cumulative histogram means for calculating a cumulative histogram, a two-dimensional histogram difference calculating means for calculating a difference in density and a differential magnitude of these two types of two-dimensional cumulative histograms, and a two-dimensional histogram obtained by the two-dimensional histogram A multiple threshold value calculating means for obtaining a plurality of density values corresponding to the peaks, and an image using the plurality of density values as threshold values. Image Segmentation method characterized by comprising: a region dividing means for dividing the frequency, the labeling means performs labeling connected components in the respective regions. 9. An image area dividing method for dividing an input image into a plurality of areas according to density values, and a differential magnitude detecting means for detecting a differential magnitude at each pixel of the input image, Minimum value filter that applies minimum value filtering, maximum value filter that applies maximum value filtering to the input image, and two-dimensional cumulative histogram that calculates the two-dimensional cumulative histogram relating to the magnitude of differentiation and the density of the image after the minimum value filtering Means, a two-dimensional cumulative histogram means for calculating a two-dimensional cumulative histogram relating to the magnitude of the differential and the density of the image after the maximum value filtering, and a difference between the density and the differential magnitude of these two types of the two-dimensional cumulative histograms. A two-dimensional histogram difference calculation means for calculating, and a plurality of two-dimensional histogram peaks obtained by the difference calculation means. A multi-threshold value calculating means for obtaining a frequency value, an area dividing means for dividing an image into areas by using a plurality of density values as threshold values, and a labeling means for labeling connected components in each area. Region division method for images. 10. A region size division method of an image, which divides an input image into a plurality of regions according to density values, and a differential size detecting means for detecting a differential size at each pixel of the input image, and a minimum size for the input image. A minimum value filter for performing a value filtering process, a two-dimensional cumulative histogram means for calculating a two-dimensional cumulative histogram relating to the magnitude of the differential and the density of the input image, and a two-dimensional cumulative relating to the magnitude of the differential and the density of the image after the minimum value filtering. A two-dimensional cumulative histogram means for calculating a histogram, a two-dimensional histogram difference calculating means for calculating a difference between density and a differential magnitude of these two types of two-dimensional cumulative histograms, and a cumulative histogram for the density of an input image And the cumulative histogram relating to the density of the image after the minimum value filtering is calculated. A cumulative histogram means, a histogram difference calculating means for calculating a difference for each density of these two types of cumulative histograms, and a two-dimensional histogram obtained from the two-dimensional histogram difference calculating means for each density and differential magnitude. Two-dimensional histogram quotient calculating means for dividing by a value corresponding to the same density of the histogram obtained by the histogram difference calculating means, and multiple threshold value calculation for obtaining a plurality of density values corresponding to the peaks of the two-dimensional histogram obtained thereby An image area dividing method comprising: a means, an area dividing means for dividing an image into areas by using a plurality of density values as thresholds, and a labeling means for labeling connected components in each area. 11. An area dividing method of an image, which divides an input image into a plurality of areas according to density values, and a differential magnitude detecting means for detecting a differential magnitude at each pixel of the input image, and a maximum of the input image. A maximum value filter for performing a value filtering process, a two-dimensional cumulative histogram means for calculating a two-dimensional cumulative histogram relating to the magnitude of the differential and the density of the input image, and a two-dimensional cumulative relating to the magnitude of the differential and the density of the image after the maximum value filtering A two-dimensional cumulative histogram means for calculating a histogram, a two-dimensional histogram difference calculating means for calculating a difference between density and a differential magnitude of these two types of two-dimensional cumulative histograms, and a cumulative histogram for the density of an input image And the cumulative histogram relating to the density of the image after maximum value filtering is calculated. A cumulative histogram means, a histogram difference calculating means for calculating a difference for each density of these two types of cumulative histograms, and a two-dimensional histogram obtained from the two-dimensional histogram difference calculating means for each density and differential magnitude. Two-dimensional histogram quotient calculating means for dividing by a value corresponding to the same density of the histogram obtained by the histogram difference calculating means, and multiple threshold value calculation for obtaining a plurality of density values corresponding to the peaks of the two-dimensional histogram obtained thereby An image area dividing method comprising: a means, an area dividing means for dividing an image into areas by using a plurality of density values as thresholds, and a labeling means for labeling connected components in each area. 12. A region size division method of an image, which divides an input image into a plurality of regions according to a density value, and a differential magnitude detecting means for detecting a differential magnitude at each pixel of the input image, and a minimum value for the input image. A minimum value filter for performing a value filtering process, a maximum value filter for performing a maximum value filtering process on an input image, and a two-dimensional cumulative histogram means for calculating a two-dimensional cumulative histogram relating to the magnitude of differentiation and the density of the image after the minimum value filtering process. And a two-dimensional cumulative histogram means for calculating a two-dimensional cumulative histogram relating to the magnitude of the differential and the density of the image after the maximum value filtering, and the difference between the density and the differential magnitude of these two types of the two-dimensional cumulative histograms. A two-dimensional histogram difference calculation means and a cumulative histogram relating to the density of the image after the minimum value filter processing A cumulative histogram means for calculating a density histogram of an image after maximum value filtering, a histogram difference calculating means for calculating a difference for each density of these two types of cumulative histograms, and the two-dimensional A two-dimensional histogram quotient calculation unit that divides the two-dimensional histogram obtained from the histogram difference calculation unit by a value corresponding to the same concentration of the histogram obtained by the histogram difference calculation unit for each density and differential magnitude, and Multiple threshold calculation means for obtaining a plurality of density values corresponding to the peaks of the obtained two-dimensional histogram, area dividing means for dividing the image into areas using the plurality of density values as thresholds, and a label of the connected component in each area. Image region segmentation method characterized by including a labeling means for performing labeling . 13. An image area dividing method for dividing an input image into a plurality of areas according to density values, and a differential magnitude detecting means for detecting a differential magnitude at each pixel of the input image; A minimum value filter for performing a value filtering process, a two-dimensional cumulative histogram means for calculating a two-dimensional cumulative histogram relating to the magnitude of the differential and the density of the input image, and a two-dimensional cumulative relating to the magnitude of the differential and the density of the image after the minimum value filtering. A two-dimensional cumulative histogram means for calculating a histogram, a two-dimensional histogram difference calculating means for calculating a difference between density and a differential magnitude of these two types of two-dimensional cumulative histograms, and a cumulative histogram for the density of an input image And the cumulative histogram relating to the density of the image after the minimum value filtering is calculated. A cumulative histogram means, a two-dimensional histogram difference calculating means for calculating the density difference between these two types of cumulative histograms, a cumulative histogram means for calculating a cumulative histogram relating to the magnitude of differentiation, and a two-dimensional histogram difference calculating means. The two-dimensional histogram obtained from the two-dimensional histogram, the value corresponding to the same density of the histogram obtained by the two-dimensional histogram difference calculation means and the same differential magnitude of the histogram obtained from the cumulative histogram means. The quotient calculation means for dividing the two-dimensional histogram by the sum of the two values, and the multiple threshold calculation means for obtaining a plurality of density values corresponding to the peaks of the two-dimensional histogram obtained by the quotient. Area dividing means for dividing the image into areas with the value as a threshold, and for each area An image area dividing method, comprising: a labeling means for labeling connected components. 14. An image area dividing method for dividing an input image into a plurality of areas according to density values, and a differential magnitude detecting means for detecting a differential magnitude at each pixel of the input image, and a maximum of the input image. A maximum value filter for performing a value filtering process, a two-dimensional cumulative histogram means for calculating a two-dimensional cumulative histogram relating to the magnitude of the differential and the density of the input image, and a two-dimensional cumulative relating to the magnitude of the differential and the density of the image after the maximum value filtering A two-dimensional cumulative histogram means for calculating a histogram, a two-dimensional histogram difference calculating means for calculating a difference between density and a differential magnitude of these two types of two-dimensional cumulative histograms, and a cumulative histogram for the density of an input image And the cumulative histogram relating to the density of the image after maximum value filtering is calculated. A cumulative histogram means for calculating a density difference between these two types of cumulative histograms, a cumulative histogram means for calculating a cumulative histogram relating to the magnitude of differentiation, and a difference calculation means for the two-dimensional histogram. The two-dimensional histogram is set to the value corresponding to the same density of the histogram obtained from the difference calculating means of the two-dimensional histogram for each density and the magnitude of the differential and the same differential magnitude of the histogram obtained from the cumulative histogram means. Two-dimensional histogram quotient calculating means for dividing by the sum of the corresponding values, multiple threshold value calculating means for obtaining a plurality of density values corresponding to the peaks of the two-dimensional histogram obtained by this, and these plurality of density values Area dividing means for dividing the image into areas as a threshold, and connection in each area An image area dividing method comprising a labeling means for labeling components. 15. A region size division method of an image, which divides an input image into a plurality of regions according to density values, and a differential magnitude detecting means for detecting a differential magnitude at each pixel of the input image, and a minimum value for the input image. A minimum value filter for performing a value filtering process, a maximum value filter for performing a maximum value filtering process on an input image, and a two-dimensional cumulative histogram means for calculating a two-dimensional cumulative histogram relating to the magnitude of differentiation and the density of the image after the minimum value filtering process. And a two-dimensional cumulative histogram means for calculating a two-dimensional cumulative histogram relating to the magnitude of the differential and the density of the image after the maximum value filtering, and the difference between the density and the differential magnitude of these two types of the two-dimensional cumulative histograms. And a cumulative histogram for calculating a cumulative histogram relating to the density of the input image. Means, a cumulative histogram means for calculating a cumulative histogram relating to the density of the image after the maximum value filtering processing, a two-dimensional histogram difference calculating means for computing a density difference between these two types of cumulative histograms, and a cumulative value relating to the magnitude of differentiation. The cumulative histogram means for calculating a histogram and the two-dimensional histogram obtained from the difference calculation means of the two-dimensional histogram correspond to the same density of the histogram obtained from the difference calculation means of the two-dimensional histogram for each density and differential magnitude. And a quotient calculating means for a two-dimensional histogram that divides by a sum of a value corresponding to the same differential magnitude of the histogram obtained from the cumulative histogram means, and a plurality of quotients corresponding to the peaks of the two-dimensional histogram obtained thereby. A multiple threshold value calculating means for obtaining each density value, and a plurality of these density values. An image region dividing method comprising: a region dividing unit that divides an image into regions using a frequency value as a threshold; and a labeling unit (L) that labels connected components in each region. 16. The claim 10, the claim 11, the claim 12, the claim 13, the claim 14, or the claim 15.
The density of the input image described in, or the density of the image after the minimum value filtering, or the cumulative histogram of the density of the image after the maximum value filtering, instead of the magnitude of the differential and the density of the input image or the differential Characterized by using a partial histogram in which the magnitude of the differential is zero in the two-dimensional cumulative histogram of the magnitude and the density of the image after the minimum value filtering, or the magnitude of the differential and the density of the image after the maximum value filtering Region segmentation method for captured images. 17. The peak of the two-dimensional histogram according to claim 7 or claim 8 or claim 9 or claim 10 or claim 11 or claim 12 or claim 13 or claim 14 or claim 15. An image region segmentation method characterized in that, as means for obtaining a plurality of corresponding density values, means for obtaining a plurality of density values corresponding to a ridge parallel to an axis relating to the degree of differentiation of a two-dimensional histogram is used. 18. The cumulative histogram relating to the magnitude of the differential according to claim 13, 14 or 15, or 16 is a two-dimensional cumulative histogram relating to the magnitude of the differential and the density of the input image. An image region segmentation method characterized by using a partial histogram of zero. 19. A two-dimensional histogram difference calculating means for each density and differential magnitude of the two-dimensional histogram obtained from the two-dimensional histogram difference calculating means according to claim 13 or 14 or 15. In a quotient calculation means of a two-dimensional histogram, which is divided by a sum of a value corresponding to the same density of the histogram obtained from the above and a value corresponding to the same differential magnitude of the histogram obtained from the cumulative histogram means, the sum is weighted linearly. An image region segmentation method characterized by summing.