JPH06101032B2 - Image processing method - Google Patents

Description

Description: TECHNICAL FIELD OF THE INVENTION Pixel values that are not to be processed are determined in advance as surrounding pixel values, and a figure that is not a surrounding pixel in an image is continuously formed with the shape features of the figure remaining. Pixels that are judged to be a 1-pixel-width curved line and are not end points or independent points, and are judged to be a sequential type, and the number of connections is 1 and are judged to be a boundary pixel if judged to be a parallel type so as to be converted into a curve of 1 pixel width. The present invention relates to an image processing method for changing a pixel value to a pixel value around the figure.

[Background of the invention and its problems]

For example, in a microscopic image of a crystal structure, the crystal interface appears as a curve having a certain width, and in analyzing the crystal structure, if the curve is not corrected to a curve having a width of one pixel passing through the center, the grain size and the like should be measured. I can't. At this time, it is necessary to thin the curve of the interface to one pixel width. If the curve of the interface becomes discontinuous during thinning, adjacent crystal grains will be connected and accurate particle size measurement cannot be performed. Based on this need, there is an almost established demand for thinning. That is, (1) the line width is 1.

(2) The position of the line is almost at the center of the original figure.

(3) The connectivity of the figures is preserved.

(4) The end portion of the line figure does not shrink more than necessary.

(5) No whiskers occur due to the unevenness of the boundary.

(6) Lines are not distorted at the intersection of figures.

Among these conditions, the conditions (1) to (4) are called basic conditions for thinning processing, and when configuring an algorithm,
Must be considered. (Junichi Hasegawa et al .: Basic technique of image processing, Technical Review, 1961, p66-67) As the image processing method for thinning, the HILDITCH method has been known so far, and the processing result is highly evaluated. there were. However, the HILDITCH method has a problem that the processing speed is insufficient because not only the number of connected pixels of the target pixel but also the number of connected pixels of the surrounding pixels are evaluated.

[Object of the Invention]

The present invention was devised to solve such conventional problems, and an object of the present invention is to provide an image processing method in which the processing result is good and the processing speed is high.

[Outline of Invention]

The image processing method according to the present invention uses the end point processing as a parallel type to prevent "beards" from remaining behind in the scanning direction, and further, to prevent the "whiskers" from remaining behind in the scanning direction.
Paying attention to the processing of the straight line of the pixel width, the pixel value of the target pixel is set as the index pixel value when there is a possibility that a straight line of the two pixel width occurs, and the index pixel The pixels adjacent to the value pixel are left, and the pixel having the index pixel value is changed to the surrounding pixel value at the next scanning.

Example of Invention

Next, an embodiment of an image processing method according to the present invention will be described with reference to the drawings.

As shown in FIG. 1, a point A (independent point) of one pixel in the image,
When there is a straight line L having a width of one pixel, if these lines are deleted, the characteristics of the image will be damaged. The end point LE of the straight line L
If 1 or LE2 is erased, the other points are also erased sequentially, and finally the straight line L is erased, and the characteristics of the image are impaired. Therefore, independent points and end points should not be deleted in the thinning process.

Further, when a certain pixel is required and two or more figures are connected, if that pixel is deleted, the image topologically changes. This connection state is evaluated by the number of connections for eight connections, and the number of connection of pixels that are the core of two or more figures is two or more. Therefore, in order to preserve the features of the image, pixels with a concatenation number of 1 should not be erased.

In the 3 × 3 convolution shown in FIG. 13, if the pixel value of the pixel to the right of the target pixel P is PO and the pixel values of the surrounding pixels are P1 to P7 in the counterclockwise direction, the number of connections
N8 is given by the following equation.

_k = 1-P _k On the other hand, in Fig. 2, when a square of a specific pixel value (for example, "1") exists in the image, in order to thin the square, the boundary pixel (Fig. (Indicated by a cross mark.)
Is changed to a peripheral pixel value (for example, “0”), and the boundary pixels are sequentially changed from “1” to “0” in each scanning. At this time, if the scanning direction is normal scanning from left to right and from top to bottom, the lower right pixel in the original square, that is, the pixel scanned last in the figure has other boundary pixels surrounding pixel values. When it changes to (marked with x), it becomes an end point and cannot be erased. Similarly, the upper left pixel and the upper left (center) pixel remain in each scanning, and finally a downward whisker "beard" as shown in FIG. 2 is generated.

In order to prevent the occurrence of such a “whisker”, the end points are processed in parallel here. Here, the parallel type processing is to judge that a pixel whose pixel value has changed a single time in one scanning is regarded as having not changed. That is, in FIG. 2, when the pixel marked with the changed X is not changed, pixels A and B having the same pixel value exist above and to the left of the pixel E, and E is not an end point. Therefore, E changes to the surrounding pixel value in the first scanning, and the upper left pixel also changes to the surrounding pixel value in the next scanning, thus preventing the occurrence of "whiskers".

Sequential processing is a processing method in contrast to parallel processing, and in this case, a pixel whose pixel value has changed in one scan is evaluated based on the pixel value after the change.

However, when the end points are processed in parallel as described above, the following problems occur. That is, a straight line having a width of 2 pixels along the traveling direction of the scanning line as shown in FIG. 4A or a straight line having a width of 2 pixels along the parallel direction of the scanning lines as shown in FIG. It disappears when processed in parallel.

In FIG. 4 (a), the pixel A at the upper left end is not an end point or an independent point, the number of connections is 1 and it is a boundary pixel, so it can be changed to the surrounding pixel value (for example, "O"). If the pixel A does not exist, the pixel B on the right side of the pixel B is left as an end point, but since the processing regarding the end point is parallel, the processing regarding the end point of B is performed assuming that A remains. Therefore, the pixel B is erased. Next, regarding the second-stage pixels C and D that have advanced by one scan in the scanning line parallel direction,
First, C is erased because it is merely a boundary pixel, and further, since the determination regarding the boundary pixel of pixel D is made assuming that A, B, and C exist, pixel D is also erased. In this way, the vertical line having a width of 2 pixels disappears at last.

In FIG. 4 (b), the first-stage pixels A, A, ... Are simply boundary pixels, and are sequentially erased. When the second-stage pixel determination is performed, the second-stage leftmost pixel B is If the pixels in the first row are erased, they are left as endpoints, but since the endpoints are determined in parallel, pixel B is erased. Similarly, B'to the right of B and its neighbor to the right are sequentially erased, and finally, all horizontal lines having a width of 2 pixels disappear.

In order to prevent the disappearance of such a straight line having a width of 2 pixels, the following conditions ai) to av) are adopted.

That is, in the 3 × 3 convolution of FIGS. 5 to 9, ai) When the upper and left pixels of the target pixel A have the surrounding pixel values (for example, “O”), the target pixel A is usually An index pixel value (a pixel value that cannot exist in the image, for example, "-1") different from the pixel value of is given (FIG. 5).

a-ii) When the condition of ai) is not met, the pixel C above the target pixel A is the index pixel value, and the pixel B below is the surrounding pixel value, determined by the sequential type. At this time, the target pixel A is saved as it is (FIG. 6).

a-iii) When the condition of ai) is not met, the pixel C on the left side of the target pixel A is the index pixel value, and the pixel B on the right side is the surrounding pixel value, which is determined by the sequential type. At this time, the target pixel A is saved as it is (FIG. 7).

a-iv) ai) to a-iii) are not met, and the judgment is performed sequentially, and the pixel B above the target pixel A is an index pixel value, and the judgment is performed in parallel. , When the pixels C ₁ and C _{2 on} the left and right of the pixel B have peripheral pixel values, the target pixel A is set as an index pixel value.

av) ai) to a-iii) are not met, the pixel B on the left of the target pixel A is an index pixel value, which is determined by the sequential type, and the pixel B is determined by the parallel type. When the pixels C ₁ and C ₂ above and below B have peripheral pixel values, the target pixel A
Is an index pixel value.

Then, the pixel having the index pixel value generated by such processing is set to the pixel value of the surroundings in the next scanning.

Unless the scanning direction is left to right or top to bottom, these conditions ai) to av) are generalized as bi) to bv) as follows.

bi) Judging by the parallel type, the pixel one pixel before the target pixel in the traveling direction of one scanning line and the pixel corresponding to the position of the target pixel in the scanning line one previous in the scanning line parallel direction are the surrounding pixels. If it is a pixel value or an index pixel value, the target pixel is converted into an index pixel value.

b-ii) When the condition of bi) is not met, the pixel corresponding to the position of the target pixel in the scanning line one before in the parallel direction of the scanning lines is the index pixel value, which is determined by the sequential type. Yes, and when the pixel corresponding to the position of the target pixel on the scanning line one after in the parallel direction of the scanning lines has the surrounding pixel value, the target pixel is stored as it is.

In the case where the condition of b-iii) bi) is not met, it is judged by the sequential type that the pixel one pixel before is the index pixel value and the pixel one pixel after is one pixel in the parallel direction of one scanning line. When the pixel value is the surrounding pixel value, the target pixel is saved as it is.

b-vi) bi) to b-iii) are not met, the pixel corresponding to the position of the target pixel in the scanning line immediately before in the parallel direction of the scanning lines is determined by the sequential type. When the pixel value is an index pixel value and the pixels before and after this pixel are surrounding pixel values in the advancing direction of one scanning line determined by the parallel type, the target pixel is set as the index pixel value.

bv) bi) to b-iii) are not met, the judgment is made sequentially, and 1 in the traveling direction of one scanning line
If the preceding pixel is an index pixel value and is judged in parallel type, and the pixel corresponding to the position of this pixel in the scanning line before and after this pixel in the parallel direction of the scanning line is the surrounding pixel value, Let the pixel be an index pixel value.

These conditions correspond to the processing of the graphics of Figures 10-12.

The figure in FIG. 10 (a) is a vertical straight line having a width of 2 pixels and corresponds to the conditions ai) and a-iii). In this figure, the pixel on the upper left side is the object of judgment, and this pixel is given an index pixel value (for example, "-1") according to the condition of ai). Next, the pixel to the right of the
According to a-iii), the pixel value (for example, “1”) as it is is not left as the surrounding pixel value. As a result, the vertical straight line having a width of 2 pixels is thinned to a straight line having a width of 1 pixel on the right side except for the pixel at the lowermost end.

The figure in FIG. 10 (b) is a horizontal straight line having a width of 2 pixels and corresponds to the conditions ai) and a-ii). In this figure, the pixel at the left end of the upper row is the target of judgment, and a-
An index pixel value (for example, "-1") is added according to the condition of i). In the upper row, all pixels except the leftmost pixel are erased. Next, when moving to the second stage, the pixel value at the left end is left as it is under the condition a-ii) (for example, “1”), and the subsequent pixels are also left in the same manner. As a result, a horizontal straight line having a width of 2 pixels is left as a straight line having a width of 1 pixel in the lower stage except for the rightmost pixel.

The figure in FIG. 11 (a) is obtained by adding a pixel A to the left at the upper end of a vertical straight line having a width of 2 pixels, and corresponds to the conditions ai) and a-iv). (However, a graphic with pixels added to the right is also processed in the same manner.) If a simple parallel type end point determination condition is used for this graphic, first, pixel A is not an end point (contacts two pixels B and C). .) So it will be converted to surrounding pixel values. Next, when moving to a portion of a width of 2 pixels, the pixel B on the left side is determined as the pixel A on the upper side is present, and is therefore converted into the pixel value of the surrounding pixels. Is converted to a value. Therefore, this figure finally disappears.

Therefore, as shown in FIG. 12 (a), the index pixel value is given to the pixel A under the condition ai) and the index pixel value is given to the pixel B also under the condition a-iv). As a result, this figure is thinned into a straight line having a width of one pixel on the right side except for the pixel at the bottom end.

The figure in FIG. 11 (b) is one in which a pixel A is added to the upper left end of a horizontal straight line having a width of 2 pixels.
-v). (However, a graphic with pixels added to the lower side is processed in the same manner.) If a simple parallel type end point judgment condition is used for this graphic, first, pixel A is not an end point (contacts two pixels B and C). It will be converted to the surrounding pixel values. Next, when moving to the 2-pixel width portion, the upper pixel B is determined as the pixel A on the left side is present, and is therefore converted into the peripheral pixel value, and finally the upper pixels are all peripheral pixels. Is converted into a pixel value of. Next, when moving to the second stage, the pixel at the left end is erased due to the presence of the pixel A, and finally the pixels in the second stage are all converted to the surrounding pixel values. Therefore, this figure eventually disappears.

Therefore, as shown in FIG. 12B, the index pixel value is given to the pixel A under the condition ai) and the index pixel value is given to the pixel B under the condition av). As a result, this figure is thinned to a straight line having a width of one pixel on the lower side except for the pixel at the right end.

In this way, a straight line having a width of 2 pixels is surely left as a straight line having a width of 1 pixel without causing a "whisker", and thus the thinning processing result is good. Since the processing of the boundary pixels of one layer is completed by one scan, the processing is fast.

Here is the data that the simulation result by the BASIC program was equivalent to or better than the HILDICH method in less than half the processing time.

In the above embodiments, the case where there is a figure of one kind of pixel value other than the surrounding pixel values in the image has been described.
Various pixel values, for example, when labeling numbers are assigned to each group of figures, each labeling number is treated as a pixel value to be processed, and each group is thinned as described above or a desired labeling number It goes without saying that only the figure can be thinned in the same manner as described above.

〔The invention's effect〕

As described above, the image processing method according to the present invention uses the end point processing as a parallel type to prevent a "whisker" from remaining behind in the scanning direction, and further, to meet the scanning line advancing direction and the scanning line paralleling method. Pay particular attention to the processing of a straight line with a pixel width, and when there is a possibility that a straight line with a two-pixel width occurs, the pixel value of the target pixel is set as an index pixel value, and when it is confirmed that the pixel width is two The pixel adjacent to the pixel having the pixel value is left, and the pixel having the index pixel value is changed to the surrounding pixel values during the subsequent scanning, so that the processing result is good and the processing is fast. It has an excellent effect.

[Brief description of drawings]

FIG. 1 is a conceptual diagram showing end points and independent points, FIG. 2 and FIG. 3 are conceptual diagrams showing the occurrence of “whiskers”, and FIG. 4 (a) is a conceptual diagram showing vertical lines of 2 pixel width, FIG. 4 (b) is a conceptual diagram showing a horizontal line having a width of 2 pixels, and FIG. 5 is a first diagram for a 3 × 3 region.
6 to 9 are conceptual diagrams showing the judgment conditions of
A conceptual diagram showing the fifth judgment condition, FIG. 10 (a) is a conceptual diagram showing a processing result of a vertical line of 2 pixel width, and FIG. 10 (b) is a conceptual diagram showing a processing result of a horizontal line of 2 pixel width. , Fig. 11 (a) is 2
11 is a conceptual diagram showing a figure in which one pixel is added to a vertical line having a pixel width.
FIG. 12B is a conceptual diagram showing a figure in which one pixel is added to a horizontal line having a width of 2 pixels, FIG. 12A is a conceptual diagram showing a processing result of the figure in FIG. 11A, and FIG. 12B. Is a conceptual diagram showing the processing result of the graphic of FIG. 11 (b), and FIG. 13 is a conceptual diagram showing a 3 × 3 convolution for explaining the number of connections. A, B, C, C ₁ , C ₂ , A ', B' ... Pixel L ... Straight lines LE1, LE2 ... End points. P _{0 to} P ₇ ... Pixel value

Claims (1)

[Claims] 1. A pixel which is not a processing target is defined in advance as a surrounding pixel value, and a figure having no surrounding pixel value in the image is converted into a continuous 1-pixel-wide curve while leaving the shape feature of the figure. As described above, it is determined by the parallel type that it is not an end point or an independent point, the sequential type is determined, and the number of connections is 1 and the parallel type is determined, and a pixel that is a boundary pixel is a pixel value around the figure. In the image processing method of changing to, the index pixel value that cannot exist in the image is defined, the pixels in the image are sequentially scanned to be the target pixel, and when the target pixel is the index pixel value, An image processing method, wherein a pixel is set as a surrounding pixel value, and the following processing is performed when it is not a surrounding pixel value or an index pixel value. i) Judging by the parallel type, the pixel preceding the target pixel in the traveling direction of one scanning line and the scanning line parallel direction is 1
When the pixel corresponding to the position of the target pixel on the immediately preceding scanning line is the surrounding pixel value or the index pixel, the target pixel is converted to the index pixel value. ii) In the case where the condition of i) is not met, the pixel corresponding to the position of the target pixel in one previous scan line in the parallel direction of the scan lines is the index pixel value, which is determined by the sequential type. In addition, when the pixel corresponding to the position of the target pixel on the next scanning line in the parallel direction of the scanning lines has the peripheral pixel value, the target pixel is stored as it is. iii) In the case where the condition of i) is not met, the pixel that is one pixel before in the traveling direction of one scanning line is the index pixel value, and the pixel that is one pixel after is one pixel in the surrounding direction when judged by the sequential type. If it is a value, the target pixel is saved as it is. iv) In the case where the conditions of i) to iii) are not met, the pixel corresponding to the position of the target pixel on the scan line immediately before in the parallel direction of the scan lines is determined as the index pixel value in the sequential determination. Further, when the pixels before and after this pixel in the advancing direction of one scanning line are the peripheral pixel values in the advancing direction of one scanning line, the target pixel is set as the index pixel value. v) In the case where the conditions of i) to iii) are not met, it is judged by the sequential type, and the pixel immediately before in the traveling direction of one scanning line is the index pixel value, and judged by the parallel type. , When the pixel corresponding to the position of this pixel in the scanning line before and after this pixel in the parallel direction of the scanning lines has the surrounding pixel value, the target pixel is set as the index pixel value. vi) When the conditions of i) to v) are not met, the target pixel is set as the surrounding pixel value.