JP2856989B2 - Inner / outer loop identification method for region boundaries - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for identifying whether a boundary of a region on a digital image is an inner loop or an outer loop.

[0002]

2. Description of the Related Art As described on pages 67 to 68 of Makoto Nagao, "Image Recognition Theory", a region code is defined by a chain code. , The sequence {U _{i + 1/2} } of the chain code difference between adjacent boundary points is expressed as U _{i + 1/2} = a _{i + 1} −a _i + 8c, where i = 1, 2,. . . , M. c is U _{i + 1/2}
Is set to one of 0, +1, and -1 so that the absolute value of is equal to or less than 4. If the region boundary is a closed loop, It is known that when this value is +8, it is counterclockwise, and when this value is -8, it is known that it is clockwise.

[0003] Focusing on this, if the area boundary is tracked so that the inside of the area is viewed on the left hand, the above-mentioned ΔU
When the value of _{i + 1/2} is +8, the region is determined to be an outer loop because the region is in a counterclockwise direction. When the value of _{i + 1/2} is −8, the region is determined to be an inner loop since the region is clockwise. Conventionally, the most common method is such a method of tracking a boundary and a method of identifying an inner / outer loop of an area boundary based on a direction around the loop.

As shown in Japanese Patent Application Laid-Open No. 3-218567 (particularly, FIGS. 17 and 18 of the publication), the object is limited to a binary image, and the first boundary point for tracking is defined as follows. A method is also known in which mask patterns for outer loop detection and inner loop detection are set in advance, and only patterns that match the respective mask patterns are sequentially tracked to the outside and the inside.

[0005]

However, the first conventional method has a problem in that the calculation for obtaining the difference between the chain codes is required at all points, so that the processing amount is large and high-speed processing is difficult. is there. Further, the second conventional method requires two times of image scanning, one for the outer loop detection and one for the inner loop detection, so that high-speed processing is difficult. There is a problem that the inner / outer loop cannot be determined for the boundary. Further, the second conventional method is not a method applicable to general images because the object is limited to a binary image.

The present invention solves the above-mentioned problems, has a small amount of processing, facilitates high-speed processing, and has a processing target of 2
It is an object of the present invention to provide a method for identifying an inner / outer loop of an area boundary applicable to digital images in general without being limited to value images.

[0007]

SUMMARY OF THE INVENTION The present invention is directed to a pixel (a top left image) at a specific position in a boundary of a region on a digital image.
Element) , the boundary point at this pixel and the boundary
The main point is to determine whether the boundary is an inner loop or an outer loop based on a local connection relationship with a boundary point.

Here, the region means a connected component of pixels having the same pixel value or the same label in the digital image. The black pixel connected component of the binary image corresponds to the special case. The term "boundary region" has two main meanings. One is a case where a pixel included in the region is set as a boundary, and the other is a case where a temporary point is regarded as a pixel boundary and this is set as a boundary. In this specification, the former is adopted. However, when the line width is 1, it is assumed that a plurality of boundary points exist in one pixel, which is substantially the same as the latter.

[0009]

According to the present invention , the leftmost pixel (called the uppermost left pixel) among the boundary pixels in the uppermost row in the region boundary
Focus on the boundary point located at. The state near the top left pixel is shown in FIG. 2 for the outer loop and the inner loop.
(A) and (b).

[0010] In order for the area boundary as shown in (a) to be an inner loop, another boundary point belonging to the same boundary exists at least above the boundary point corresponding to the uppermost left pixel. Must be done because this violates the assumption of the top left pixel. And if the region boundary is always closed and cannot be an inner loop,
It can be said that this boundary region is always an outer loop. (B)
It can be similarly explained that the region boundary shown in FIG.

That is, from the local connection relation of the boundary point at the position of the uppermost left pixel in the region boundary or the pixel pattern in the vicinity, it is determined which of FIG. 2A or FIG. That is, it can be determined whether the region boundary is an outer loop or an inner loop. This is the principle of the method of the present invention.

In the method of the present invention, the amount of calculation for inner / outer loop identification is much smaller than that of the conventional method in which it is necessary to obtain chain code differences at all points, and high-speed processing is achieved. It is possible. Further, the processing target is not limited to the area boundary on the binary image.

[0013]

Embodiments of the present invention will be described below with reference to the drawings. In the following description, as described with reference to FIG. 2, the position of the uppermost left pixel of the region boundary is selected as the specific position in the region boundary of interest.

FIG. 1 shows an example of a processing system for implementing the present invention. 1 is a central processing unit (CPU), 2 is an image memory for storing digital image data, 3 is a program memory for storing a program of the CPU 1, and 4 is a data memory for storing data such as processing results. The programs stored in the program memory 3 include a boundary extraction processing program 5 for extracting an area boundary from digital image data in the image memory 2 and boundary point information 6 (data (Stored in the memory 4) for performing an inner / outer loop identification process. Note that the inner / outer loop identification program 7 may be integrated into the boundary extraction processing program 5 in some cases. The processing described below is executed by the CPU 1 according to the corresponding program.

Next, the inner / outer loop identification processing in one embodiment will be described. The inner / outer loop identification conditions will be described more specifically.

The coordinate system is defined as shown in FIG. 4, and the boundary tracing at the time of boundary extraction is performed such that the inside of the region is viewed on the left hand or in an equivalent form. And
Let C be the boundary of interest.

It is checked how many boundary points of C exist at the position of the upper left pixel of C. When C is a line segment having a width of one pixel, up to two boundary points of C may exist at the position of the uppermost left pixel of C (see FIG. 3). In this case, C is uniquely determined to be the outer loop. When there is only one boundary point of C at the position of the top left pixel of C, the boundary point is set to p and the following processing is performed. During processing, the next symbol x (p): x coordinate of p y (p): y coordinate of p prev: one before (back) of p in the boundary tracking direction
Next: boundary point one point ahead (forward) of p in the boundary tracking direction x (prev): x coordinate of prev y (prev): y coordinate of prev x (next): x coordinate of next y (next) ): The y coordinate of next is used.

First, consider the case of 8-connection (however, in the case of an isolated point, the outer loop is determined). Here, since it is assumed that p is located at the position of the upper left pixel, in the 3 × 3 pixel area near p shown in FIG.
A boundary point connected to p cannot exist at the position of b, c, and d, and a position where a boundary point connected to p can exist is any of e, f, g, and h. That is, the connection status of the boundary points in the vicinity of p is expressed in the order of boundary tracking, that is, by prev → p → next, {e, f, g, h} → p → {e, f, g, h} Is limited to Table 1 shows the connection status when the boundary C is the outer loop and the inner loop.

[0019]

[Table 1]

In Table 1, the left column shows pre connected to p.
The positions of v and next are shown, and the right column shows the inner / outer loop. However, X in the right column means that the combination of connections shown in the left column is not possible. From Table 1, it can be seen that the inner loop can be limited to only f → p → e and f → p → h.

If the symbols defined above are used, "x (prev) = x (p) -1 and y (pre
v) = y (p) +1 and x (next) = x (p) +1
And y (next) = y (p) or y
(Next) = y (p) +1. Is the condition for the inner loop. Here, since it is not necessary to consider the connection between c and p, it is easier to write “x (prev) = x (p) −1 and y (pre
v) = y (p) +1 and x (next) = x (p) +
It is an inner loop only when 1 is satisfied, otherwise it is an outer loop. ". Alternatively, "if the vicinity of p matches one of the mask patterns shown in FIGS. 5A and 5B, it is an inner loop; otherwise, it is an outer loop."

Next, the same condition is considered for the case of <4 connection (in the case of an isolated point, the outer loop is determined). Due to the four-connection and the assumption that p is at the position of the upper left pixel, in FIG. 5, the position where the boundary point connected to p can exist is either e or g.

That is, the connection status of the boundary point near p is limited to {e, g} → p → {e, g}. Table 2 shows the connection status of the outer loop and the inner loop.

[0024]

[Table 2]

As can be seen from Table 2, an inner loop can be limited to one of g → p → e.

If the symbols defined above are used, "x (prev) = x (p) and y (prev) = y
The loop is an inner loop only when (p) +1 and x (next) = x (p) +1 and y (next) = y (p), and the other loop is an outer loop. It can be said that this is the condition. Since it is not necessary to consider the connection between c and h and p, it is easier to write “x (prev) = x (p) and y (prev) = y
It is an inner loop only when (p) +1 and x (next) = x (p) +1 are satisfied, and the other loops are outer loops. ". Alternatively, "if the vicinity of p coincides with the mask pattern of FIG. 6C, the loop is an inner loop, and the rest is an outer loop."

With the above-described method, the inner / outer loop is identified based on the local connection state of the boundary point at the position of the uppermost left pixel of the focused area boundary or the neighboring pixel pattern. FIG. 7 is a flowchart showing the identification process. First, the number n of boundary points at the upper left pixel is obtained (step 100). This can be generally determined by comparing the boundary point coordinate values. And n is 1
Is checked (step 105), and n１０５1 (n = 1)
If 2), it is determined that the loop is an outer loop (step 11).
0), the process ends.

If n = 1, the boundary point is set to p (step 115), and the condition regarding the local connection state or the neighboring pixel pattern is checked (step 12).
0, 125), if the condition of the inner loop is not satisfied, it is determined that the loop is an outer loop (step 110), and if the condition of the inner loop is satisfied, it is determined that the loop is an inner loop (step 130), and the process ends. .

Generally, by the method described above,
Inner / outer loop identification of a given area boundary can be performed. However, when the same identification is performed for all the region boundaries in a given image, the method of obtaining the boundary point at the uppermost left pixel (step 100 in FIG. 7) is devised by devising the method of the boundary extraction processing. Response) can be further simplified.

FIG. 9 is a flowchart showing an example of such a processing algorithm. Prior to the description of the contents of this processing, the boundary (point sequence) extraction processing will be described.

The method of the boundary extraction processing can be roughly classified into a tracking type algorithm and a scanning type algorithm. The tracking-type algorithm scans the image, finds a boundary pixel that has not been tracked yet, and if detected, extracts and marks a boundary point on the boundary, and then returns the image from the previous detection position again. This is a method of continuing scanning. The feature of this algorithm is that the boundary point information is arranged in the tracked order on the memory, starting from the boundary point of the upper left pixel.

The scanning type algorithm, which is another boundary extraction method, extracts the boundary point if there is a boundary point while scanning the image together with the connection information with the neighbors, and later (or simultaneously) extracts the boundary point based on the connection information. It is a method of connecting points. The feature is that the boundary point information is arranged in the memory in the scanning order, but the boundary point of the uppermost left pixel is recorded at the top of one boundary.

Here, considering that there may be a plurality of boundary points at the position of the uppermost left pixel, the following restrictions are applied to the above two types of boundary extraction processing. That is, when there is a pattern as shown in FIG. 8A in the case of 8-connection, and when there is a pattern as shown in FIG. Boundary points related to the performed boundary are extracted before the boundary points related to the boundary indicated by the thin arrow.

By adding such a restriction, the most important one of the boundary points at the upper left pixel is closest to the head in the memory regardless of the tracking algorithm or the scanning algorithm. It will be recorded there. In the case of the tracking type algorithm, the most important one of the boundary points of the uppermost left pixel is a tracking start target of the boundary.

Therefore, assuming this restriction, special processing for selecting the boundary point of the uppermost left pixel is performed even when there are a plurality of boundary points at the position of the uppermost left pixel of interest. It is not necessary, and it is possible to construct the processing algorithm as shown in FIG. Hereinafter, FIG.
Will be described, but it is assumed that the boundary point information has already been extracted.

First, the labels of all the boundary points are set to 0 (step 200). Next, the boundary points with the label 0 are extracted in the order of extraction, that is, in order from the top one in the memory (step 205). When there are no more boundary points with a label of 0 (step 210), the process ends.

If the boundary point where the label is 0 can be extracted,
Since this boundary point is always the boundary point of the uppermost left pixel of a certain region boundary (the most important one), it is set to p (step 215). Then, based on the inner / outer loop determination condition as described above, it is determined whether the boundary is an outer loop or an inner loop from the connection state of the boundary point near p or the pixel pattern near p ( Step 2
20-235). Then, the labels of all the boundaries belonging to this boundary are set to 1 (Step 240), and Step 2
The process is repeated from 05. When the end of the processing is determined in step 210, the distinction of the inner / outer loop is determined for all the boundaries in the image.

If a tracking type algorithm is used for boundary extraction, it is possible to perform inner / outer loop discrimination simultaneously with the boundary extraction processing. This is because the boundary point to be the tracking start target is the boundary point at the uppermost left pixel (the most important one) at the boundary. In other words, when the tracking start target is found, it is only necessary to perform the processing of identifying the inner / outer loop and then tracking the boundary while marking the boundary points belonging to the boundary.

[0039]

As is apparent from the above detailed description, according to the present invention, a boundary point located at the upper leftmost pixel of an area boundary is intended for digital images in general, not limited to binary images.
And the connection between this and the nearby boundary points.
By determining the inner / outer loop, it becomes possible to perform the inner / outer loop identification of the region boundary at a high speed with a small amount of processing.

[Brief description of the drawings]

FIG. 1 shows a configuration of an example of a processing system for implementing the present invention.

FIG. 2A shows a state near an upper left pixel of an area boundary in the case of an outer loop. (B) The state near the top left pixel of the area boundary is shown for the case of the inner loop.

FIG. 3 shows an example in which there are two boundary points at the same boundary at the position of the upper left pixel.

FIG. 4 shows a coordinate system.

FIG. 5 shows a position near a boundary point of an uppermost left pixel position.

FIG. 6A shows a mask pattern in the case of 8-connection. (B) shows a mask pattern in the case of 8-connection. (C) shows a mask pattern in the case of four connections.

FIG. 7 shows a processing algorithm for inner / outer loop identification of one boundary.

FIG. 8A shows a mask pattern in the case of 8-connection. (B) shows a mask pattern in the case of four connections.

FIG. 9 shows an algorithm of processing for performing inner / outer loop discrimination of all area boundaries in an image.

[Explanation of symbols]

 Reference Signs List 1 CPU 2 Image memory 3 Program memory 4 Data memory 5 Boundary extraction processing program 6 Boundary point information 7 Inner / outer loop discriminating program

Claims (1)

(57) [Claims] 1. A boundary of an area which is a connected component of pixels having the same pixel value or the same label on a digital image (hereinafter referred to as a boundary).
(Referred to as a region boundary) is an inner loop or an outer loop. The method according to claim 1, wherein a boundary point located at an upper left pixel in the region boundary is focused on. Characterized in that whether the area boundary is an inner loop or an outer loop is identified by the local connection relation of (1) and (2).