JPH0850655A - Edge trace method - Google Patents

Abstract

PURPOSE:To obtain a comparatively long edge chain even if edge image data is finely divided. CONSTITUTION:When a processing searching the next edge picture element existing in the range of 1 pix around an edge picture element by defining the edge picture element as a search center from edge image data is repeated by successively shifting the search center, the search of the next edge picture element is performed S 4 by shifting the search center to a discovered edge picture element S 6 when this edge picture element can be discovered. When the edge picture element can not be discovered, an edge trace is not terminated and the location of the next edge picture element is estimated from the positional data of the already determined edge picture element S 9. The processing is continued by shifting the search center to the picture element at this estimated location S 6.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an edge tracing method in so-called image processing, and more particularly to an edge chain which is a series of edge pixels which are pixels satisfying a specific condition from an arbitrary spatial image data group. It relates to an improvement for improving the trace performance when obtaining.

[0002]

2. Description of the Related Art Conventionally, in order to extract edge data of an object displayed on a screen in image processing, conventionally, the following steps (1) to (3) have been generally performed. Become. However, ~ corresponds to the edge trace.

Edge image data, that is, image data having an edge width of 1 pix is created from input image data. Note that pix represents the width of one pixel. To create the edge image data, basically, the input image data is created by performing a sharpening process using a differential spatial filter or the like to emphasize the edge portion of the object or the like. If the edge width does not reach 1 pix by the sharpening processing, the edge width is set to 1 pix by further thinning processing. See steps S60 and S61 in FIG.

Based on an appropriate threshold value, the edge pixel that is the starting point of the edge trace is detected from the edge image data. See step S62 in FIG.

With the detected edge pixel (starting point) as the search center, the range of 1 pix around is searched clockwise or counterclockwise, and an appropriate threshold value is set to determine whether or not the next edge pixel exists in this search range. Based on the edge image data, it is checked. See step S63 in FIG.

When the next edge pixel is found, the search center is shifted to the found edge pixel and the surrounding 1
The range of pix is searched clockwise or counterclockwise, and whether or not the next edge pixel is present in this search range is checked from the edge image data based on an appropriate threshold value. This procedure is repeated as long as edge pixels are found. See steps S64, S65 and S63 in FIG.

If no edge pixel is found, the search is terminated, and the data of the edge pixel found so far including the start point of the edge is used as an edge chain (also referred to as edge line segment data). Steps S64 and S6 of FIG.
See 6.

[0008]

As described above, in the conventional edge tracing method, if the next edge pixel cannot be found within the range of 1 pix around the search center, the search for the edge pixel is terminated at that point. To do. Therefore, even if the object has relatively long edges, if the edge image data is finely divided during the thinning process,
Alternatively, if the edge image data is also finely divided as a result of finely dividing the image data that has been sharpened due to noise or the like that occurs when inputting an image, one edge trace will result in a relatively short edge chain. I can only get it. This problem tends to be more noticeable as the resolution becomes higher.

In the conventional edge tracing method described above, when the search is started by starting the edge pixel 101 in the middle of the original edge chain 100 to be detected as shown in FIG. 20, the starting point 101 is shown as shown in FIG. Therefore, the original edge chain 100 is divided and only the edge chain 102 shorter than that is obtained.

That is, when edge image data 104 consisting of a pixel 103 with a diagonal line descending to the left as shown in FIG. 20 is given, when an intermediate edge pixel 101 becomes the starting point, pixels 105 and 106 in the vicinity thereof are provided. Only one of them, for example the right neighbor 106, is found as the next edge pixel. As a result, as shown in FIG. 21, the edge trace ends when the edge chain 102 from the next edge pixel 106 to the end edge pixel 107 is detected.

For the remaining edge chains 108 that have been divided, it is necessary to separately perform edge tracing.

The two edge chains 102 and 108 thus detected are both the original edge chain 10
Since the length is shorter than 0, it is possible to obtain the edge chain 100 of the original length by performing an operation of connecting the end points 101 and 105 of the both edge chains 102 and 108.

However, when there are many edge chains, there is a problem that the operation of checking whether or not the end points are close to each other is complicated and requires a large amount of calculation.

Therefore, according to the present invention, a relatively long edge chain can be easily obtained, which is hardly affected by noise generated when an image is input, or is hardly affected by subdivision of edge image data generated by thinning processing. It is an object of the present invention to provide an edge tracing method capable of achieving a long edge chain without performing a connecting operation between the end points of the edge chain.

[0015]

[Means for Solving the Problems] First to achieve the above object
In the edge tracing method of the invention described above, a pixel satisfying a specific condition is set as an edge pixel, the edge pixel is set as a search center, and the next edge pixel existing in the range of 1 pix around the edge pixel is searched for from the image data. In an edge tracing method for obtaining an edge chain by shifting and repeating: shifting the search center to this found edge pixel if the next edge pixel can be found in the search for said edge pixel; said edge If the next edge pixel cannot be found during the pixel search, the position of the next edge pixel is estimated from the already obtained edge pixel position data, and the search center is shifted to the pixel at this estimated position. This is a characteristic method. In this case, the second invention is characterized in that the next edge pixel position is estimated by linear approximation. In the third invention, the position of the current edge pixel is set to (X _i , Y _i ), the position of the jth previous edge pixel is set to (X _ij , Y _ij ), and the position of the next edge pixel is set to (X _{i + 1} , Y _{i + 1} ) is characterized in that the position data of the past n edge pixels is estimated by the following equation (3).

(Equation 3)

Further, the fourth invention is characterized in that the search for the edge pixel is terminated when the edge pixel cannot be found from the image data even after the estimation of the next edge pixel position is repeated a predetermined number of times. In the invention of (1), when the search for the edge pixel is completed, the final position of the edge chain is returned to the edge pixel that was found last.

On the other hand, in the edge tracing method of the sixth aspect of the invention for achieving the above object, a pixel satisfying a specific condition is set as an edge pixel, and the edge pixel is set as a search center, and the next edge pixel is extracted from the image data in the search range around the edge pixel. In the edge trace method for obtaining the edge chain by sequentially shifting the search center to the found edge pixel and repeating the process of searching, the search range at the same search center is
Until the next edge pixel is found, sequentially expand by 1pix from the surrounding 1pix range; m is an integer of 2 or more,
When k is an integer less than or equal to m, if an edge pixel can be found in the search range of the surrounding k pix, the found edge pixel and the edge pixel at the search center are interpolated by k-1 pixels. If no edge pixel can be found within the search range of the surrounding m pix, end the edge pixel search;
Is a method characterized by. In this case, in the seventh invention,
The position of the edge pixel which is the current search center is (X _i ,
Y _i ), and the position of the found edge pixel is (X _{i + k} , Y
_{i + k} ) and the position of the pixel to be interpolated h-th from the current search center is (Xi _{+ h} , Yi _{+ h} ), the interpolation of k-1 pixels is performed by the following equation (4). It is characterized by performing.

[Equation 4]

Further, the edge tracing method according to the eighth aspect of the invention for achieving the above-mentioned object is that a pixel satisfying a specific condition is set as an edge pixel, and the next edge pixel existing in a specific range around the edge pixel is set as a search center. In the edge trace method that obtains an edge chain from the start point to the end point by sequentially shifting the search center and repeating the process of searching from the image data: When the edge pixel at the end point is found, the edge chain obtained so far Determining whether or not there is an edge pixel of another edge chain near the start point of the above; if there is an edge pixel of the other edge chain, the start point and end point of the edge chain obtained so far are replaced. , Continuing to search for edge pixels from the new end point; if there is no edge pixel in the other edge chain, the edge image Ending the search; characterized.

A ninth invention is an application of the eighth invention to the first to fifth inventions, wherein an edge pixel is found from image data even if the estimation of the next edge pixel position is continued a predetermined number of times. If not, determine whether or not there is an edge pixel of another edge chain near the start point of the edge chain obtained so far; if an edge pixel of the other edge chain exists, Replacing the start point and the end point of the edge chain obtained up to now and continuing the search for the edge pixel from the new end point; ending the search for the edge pixel if the edge pixel of the other edge chain does not exist; Is an edge tracing method.

A tenth invention is an application of the eighth invention to the sixth to seventh inventions. When an edge pixel cannot be found in the search range of the surrounding m pix,
Determining whether or not an edge pixel of another edge chain exists near the start point of the edge chain obtained thus far; if an edge pixel of the other edge chain exists, it is obtained so far. An edge trace method, characterized in that the start point and the end point of an edge chain are exchanged, and the search for an edge pixel is continued from a new end point; if the edge pixel of the other edge chain does not exist, the search for the edge pixel is ended. Is.

[0021]

When the edge image data is subdivided due to noise or the like, even if the next edge pixel cannot be found in the range of 1 pix around the certain search center, the search center is slightly shifted or the search center is searched. If you expand the range without changing,
It is very likely that you can find edge pixels.

Therefore, when the next edge pixel cannot be found in the first invention, the position where the next edge pixel is likely to exist is estimated from the already obtained position data of the edge pixel, and the pixel at the estimated position is searched. The center is shifted to search for the next edge pixel. This makes it possible to obtain a relatively long edge chain. The estimation method may be any method, but the method based on the linear approximation as in the above equation (3) is simple. If the edge pixel cannot be found in one estimation, the estimation is repeated many times, but the edge image data is not subdivided due to noise or the like.
Since the edge may actually end, in such a case, when the edge pixel cannot be found even after the estimation is repeated a predetermined number of times, the edge pixel search is ended. In this case, it is possible to use the estimated position as the final position of the edge chain, but it is more reliable and preferable to return to the last found edge pixel.

On the other hand, in the sixth aspect, when the next edge pixel cannot be found, the search range is sequentially expanded from the range of 1 pix in the periphery to the range of m pix in the periphery without changing the search center, and the edge is expanded. When a pixel is searched and an edge pixel is found in the search range of the surrounding k pix, the middle is interpolated with k−1 pixels. This makes it possible to obtain a relatively long edge chain. When the edge pixel cannot be found even if it is expanded to the range of the surrounding m pix, the edge pixel search is ended by determining that the edge is really ended, not by subdivision of the edge pixel data due to noise or the like. The interpolation method may be any method, but the method of linear approximation as in the previous expression (4) is simple.

Each of the eighth, ninth and tenth inventions should be referred to as a folded edge trace method, and the edge trace end condition is that a pixel of another edge chain exists near the start point of the edge trace. By checking whether or not, it is determined whether or not to end. If another edge chain exists near the start point, the edge chain obtained up to now is turned over, that is, the start point and end point of the same edge chain are turned over, and then edge tracing is restarted from the new end point. To do. This makes it possible to correctly obtain a long edge chain that originally exists in an image, without being affected by the initial starting point position of the edge trace.

The eighth basic invention will be further described with reference to FIGS. 22 to 25, which are as follows (1) to (4). However, in these figures, the upper left is the origin, the horizontal axis is the X position, and the vertical axis is the Y position.
Indicates. The edge image data 1 is composed of pixels 2 that exceed a certain threshold value, and the pixels 2 are shown with a diagonal line to the left. Pixels (edge pixels) found as constituting an edge chain are shown with cross hatching.

(1) First, the upper left corner of the image is searched for the presence of an edge chain to find the first starting point 3 as shown in FIG. Then, from this start point 3 until there are no edge pixel candidates, as shown by the cross hatched lines in FIG.
The edge chain 20 is traced. (2) Next, it is checked whether or not there is a candidate for an edge pixel of another edge chain near the starting point 3 of the edge chain 20. (3) As shown in FIG. 24, if the edge pixel 23 of another edge chain 22 exists, the start point 3 and the end point 21 of the edge chain 20 obtained previously are turned over, and the start point 21
A and end point 3A. This is a set of edge chains 20 (X = X ₀ , X ₁ , ..., X _n , Y = Y ₀ , Y ₁ ,
…, Y _n | chain (X, Y)) is converted into (X = X, X, ...,
X, Y = Y, Y, ..., Y | chain (X, Y)). (4) Next, the edge chain 20 obtained in (2) above.
2 from the new end point 3A (chain (X _n , Y _n )) of FIG.
As shown in FIG. 5, a search is performed until there are no candidate pixels for the edge chain 22, and a long edge chain 25 from the start point 21A to the end point 24 is obtained.

[0027]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The first and second embodiments of the invention will be described below with reference to FIGS. The third and fourth embodiments of the invention will be described with reference to FIGS. 2, 3, 4, and 12 to 18. 1 and 10 show the flow of the processing procedure of the edge tracing method according to the first and second embodiments, respectively, and FIGS. 2 to 9 and 11 show screens for explaining the individual processing. The edge image data illustrated above is shown. 12 and 17 show the flow of the processing procedure of the edge tracing method according to the third and fourth embodiments, respectively, and FIGS. 13 to 16 and 18 show screens for explaining the individual processing. The edge image data illustrated above is shown. 2 to 9, FIG. 11 and FIGS. 13 to 16 and FIG. 18, each grid represents a pixel 1 pix, the grid at the upper left corner is the origin, the X position is on the horizontal axis, and the Y position is on the vertical axis. Is taking.

<First Embodiment: Edge Tracing Method by Estimating Using Linear Approximation> The processing shown in FIG. 1 is automatically performed by the image processing apparatus. Now, from the image data of a certain object input in step S1 of FIG.
Then, it is assumed that the edge image data 1 including the pixel 2 with the diagonal line to the left as shown in FIG. 2 is created. As described above, the image data in which the edges are emphasized is created by performing the sharpening process such as the differential spatial filter on the input image data, and the edge-enhanced image data is subjected to the thinning process as necessary to obtain the edges. Create edge image data 1 with a width of 1 pix.

First, in step S3 of FIG. 1, the edge pixel which is the starting point of the edge trace start is detected from the edge image data 1 of FIG. 2 based on an appropriate threshold value and stored.
In the example in FIG. 3A in which the edge pixel 3 indicated by the cross-hatched line is detected as the starting point, the starting point of the edge is detected by the following method examples (A) to (D). (A) The starting point of the edge is searched from the pixel at the upper left corner of the screen. (B) Search for the starting point of the edge along the right lateral direction. (C) If the start point of the edge cannot be found in (B) above, the start point of the edge is searched along the right lateral direction from the left end one pix lower. (D) The above (C) is repeated up to the lower edge of the screen until the start point of the edge can be found.

Next, in step S4 of FIG.
As shown in, with the start point 3 of the edge as the search center, the range of 1 pix around the edge is searched clockwise 4 or counterclockwise,
Based on an appropriate threshold value, the adjacent edge pixel following the edge start point 3 is searched. In FIG. 3 (b), the pixel 5 with cross hatching is the first pixel that satisfies the threshold value, and this is found and saved as an edge pixel.

After the edge pixel 5 is found, the processing proceeds from step S5 to steps S6 and S4 in FIG.
As shown in (c), after shifting the search center to the edge pixel 5, the range of 1 pix around the edge pixel 5 is searched clockwise 4 or counterclockwise, and the adjacent edge following the edge pixel 5 based on an appropriate threshold value. Search for pixels.

By repeating the above-described edge pixel search and search center shift, as shown in FIG. 4, even the edge pixels 6 with cross hatching can be found and saved.

However, in the example of FIG. 4, when the search center is shifted to the edge pixel 6, the next edge pixel cannot be found even if the range of 1 pix around it is searched.
Conventionally, the edge trace is finished at this stage.

On the other hand, in the present invention, the edge trace is not completed, and the pixel 7 is shown as a circled circle with diagonal lines in FIG.
The position where the next edge pixel is likely to exist is estimated, and it is assumed that the edge pixel has been found at that position, and the search for the edge pixel proceeds.

That is, if the next edge pixel cannot be found, the process proceeds from step S5 in FIG. 1 to step S9 via steps S7 and S8 described later to estimate the position of pixel 7 in FIG. In this embodiment, the position data of the past n edge pixels including the edge pixel 6 which is the current search center is approximated by a straight line, and the position obtained by advancing the straight line by 1 pix from the current search center is the estimated position of the pixel 7. I am trying. If a specific example of this estimation is expressed by an equation, it can be obtained by the following equation (5). However, in the equation (5), (X _i , Y _i ) is the position of the current edge pixel, (X _ij , Y _ij ) is the position of the jth previous edge pixel, and (X _{i + 1} , Y _{i + 1)} ) Is the estimated edge pixel 7
Position. n is an appropriate integer of 2 or more and is set as a fixed value.

(Equation 5)

When the position of the edge pixel 7 is estimated, the process proceeds to steps S6 and S4 in FIG. 1, and as shown in FIG.
The estimated edge pixel 7 is shifted to the search center, and the range of 1 pix around the edge pixel 7 is searched clockwise 4 or counterclockwise, and an adjacent edge pixel subsequent to the estimated edge pixel 7 is searched based on an appropriate threshold value. To do.

In the example of FIG. 6, the next edge pixel cannot be found. Therefore, again from step S5 to step S7 in FIG.
Then, the process proceeds to step S9 via S8 and S8, and a position where a next edge pixel is likely to exist, such as the pixel 8 indicated by a shaded circle shown in FIG. 7, is estimated.

In this case, in the above equation (5), (X _i ,
Y _i ) is the position of the previously estimated edge pixel 7 which is the current search center, (X _ij , Y _ij ) is the position of the edge pixel j number before the estimated edge pixel 7, and (X _{i + 1} , Y
_{i + 1} ) is the position of the edge pixel 8 estimated the second time.

Once the position of the edge pixel 8 has been estimated, FIG.
The processing proceeds to steps S6 and S4, and as shown in FIG. 8, the edge pixel 8 estimated for the second time is searched for a range of 1 pix around the search center as a shift, and is estimated based on an appropriate threshold value. The adjacent edge pixel following the edge pixel 8 is searched.

In the example of FIG. 8, the cross-hatched pixel 9 is found as an edge pixel and saved. Since the edge pixel 9 is found, the estimated edge pixel 7,
8 is also stored as edge chain data.

By repeating the above-described edge pixel search, edge pixel position estimation, and search center shift, a long edge chain 10 is finally obtained as shown in FIG.

Next, steps S7, S8 and S1 in FIG.
0 will be described.

In step S7, the number of times the edge pixel position is continuously estimated is counted, and in step S8, the estimated number of times the edge pixel position is estimated is set to a predetermined number, for example, 3
Compare with If the counted number of times of estimation is less than or equal to the predetermined number of times, the process proceeds to step S9 to estimate the position of the edge pixel, but if the edge pixel cannot be found even after the predetermined number of times, the process proceeds to step S10 to perform edge tracing. finish. However, in step S10, the position data of the edge pixel estimated after the last found edge pixel is discarded, and the final position of the edge chain is returned to the last found edge pixel. It is preferable that the above-described predetermined number of times be changeable and that the value be set to an appropriate value by looking at the edge trace result.

If the starting point of the edge is the first, since the number of data used for estimation is small for the edge pixels from the starting point of the edge to the n-th edge, the position of the edge pixel is not estimated and the surrounding 1 pix is used as usual. If no edge pixel is found in the range, the edge trace is terminated.

<Second Embodiment: Application of Folding Type Edge Tracing Method to First Embodiment> Next, FIGS.
A second embodiment of the present invention will be described with reference to FIGS. The second embodiment is an application of the folded edge tracing method to the first embodiment, and the processing procedure of FIG. 10 is different from that of FIG. 1 in steps S11 to S2.
Same, except that 4 was added. The processing shown in FIG. 10 is also automatically performed by the image processing apparatus. In addition, from the image data input in step S1 of FIG. 10, in the next step S2, the edge image data 1 having an edge width of 1 pixel, which is composed of pixels 2 with diagonally left-down diagonal lines as shown in FIG.
Is created.

First, in steps S3 to S9 of FIG. 10, the same processing as in the first embodiment is performed. The following is a brief summary. In step S3, as shown in FIG. 3, based on an appropriate threshold value, the edge pixel which is the starting point 3 of the edge trace start is detected. In step S4, the next edge pixel is searched in the range of 1 pix around this edge pixel 3 as the search center. If found, step S
The process of going from 5 to step S6, shifting the search center, returning to step S4, and further searching for the next edge pixel is repeated. In the example of FIGS. 3 and 4, the edge pixels 5 and 6
Can be found one after another. If the next edge pixel cannot be found in the surrounding 1 pix, steps S7 and S8
After proceeding to step S9, the position where the next edge pixel is likely to exist is estimated by the above equation (5), and it is assumed that the next edge pixel 7 can be found at that position as shown in FIG. The search center is shifted in S6, the process returns to step S4, and the process of searching for the next edge pixel as shown in FIG. 6 is repeated. This estimation is allowed up to a predetermined number of times, for example 3 times
The estimation counts a predetermined number of times in step S7 and compares with the predetermined number of times in step S8. In the example of FIGS. 5 to 8, the next edge pixel 9 is found the second time. As described above, in the first embodiment, the edge chain 10 from the start point 3 to the end point 21 is obtained by such processing as shown in FIG.

That is, in the first embodiment, the edge trace is ended when it is determined in step S8 that the estimated number of times exceeds the predetermined number.

On the other hand, in the second embodiment, the process proceeds from step S8 to step S11 and thereafter, and the edge tracing is further continued. Details are as follows.

That is, when it is determined in step S8 that the estimated number of times exceeds the predetermined number, the search center is shifted to the starting point 3 of the edge chain 10 obtained up to now in step S11, and in the next step S12, the search center is surrounded. An edge pixel of another edge chain 22 existing in the range of 1 pix is searched based on an appropriate threshold value. In step S13, it is determined whether or not the next edge pixel has been found. If found, the process proceeds to step S14. If not found, step S2.
After 1 and S22, the process proceeds to step S23.

In the example of FIG. 9, since an edge pixel of another edge chain 22 cannot be found in 1pix around the starting point 3,
After steps S21 and S22, the process proceeds to step S23,
The position where the next edge pixel is likely to exist is estimated by the above equation (5). It is estimated by the pixel 7 shown in FIG. Then, the search center is shifted to this pixel 7 in step S24, and then the process returns to step S12 and the surrounding 1 pix is again displayed.
Search within the range. In the example of FIG. 11A, the edge pixel 23 of another edge chain 22 can be found by the first estimation.

Similar to step S9 of FIG. 1 or FIG. 10, the number of times of estimation in step S23 is allowed up to a predetermined number of times, for example, three times, and the number of times of estimation is counted in step S21 and compared with the predetermined number of times in step S22. If the edge pixel of another edge chain cannot be found even after the predetermined number of times, the process proceeds to step S10, and the edge tracing is ended.

In the example of FIG. 11 (a), since the edge pixel 23 of another edge chain 22 can be found by the first estimation, the process proceeds from step S13 to step S14, and the edge chain 10 obtained up to now is searched. It is turned over and the starting point 3 thus far is set as a new ending point 3A, and the ending point 21 so far is set as a new starting point 21A.

Then, the process proceeds to the next step S15, and the new end point 3A is set as the search center, and the search for the edge pixel following the search is restarted. However, in the example of FIG. 11A, since the edge pixel 23 has already been found after the first estimation, the process proceeds from step S16 to step S17, the search center is shifted to this edge pixel 23, and the process proceeds to step S15. Returning to this, the edge pixels existing in the range of 1 pix around the new search center are searched based on an appropriate threshold value. By repeating this process, in the example of FIG. 11, even the edge pixel 24 is found without performing estimation as shown in (b).

After the search center is shifted to the edge pixel 24, the next edge pixel cannot be found in the surrounding 1 pix, so the process proceeds from step S16 to steps S18 and S19 to step S20, where the following equation (5) The position where the edge pixel is likely to exist is estimated, the search center is shifted to that position in step S17, and the process returns to step S15 to perform the search in the surrounding 1 pix. The number of times of estimation here is also allowed a predetermined number of times, for example, three times, and the number of times of estimation is counted in step S18 and compared with the predetermined number of times in step S19. If the next edge pixel cannot be found even after the predetermined number of times, the process proceeds to step S10, the edge tracing is ended, and the edge pixel 24 found last is set as the end point.

Thus, as shown in FIG. 11B, a long edge chain 25 from the starting point 21A to the ending point 24 is obtained.

<Third Embodiment: Edge Tracing Method for Increasing Search Range Sequentially in Waveform> The processing shown in FIG. 12 is also automatically performed by the image processing apparatus. In addition, from the image data input in step S30 of FIG.
It is assumed that the edge image data 1 having an edge width of 1 pix composed of the pixel 2 with the diagonal line to the left as shown in FIG.

First, also in step S32 of FIG. 12, the edge pixel which is the starting point of the edge is detected from the edge image data 1 of FIG. 2 based on an appropriate threshold value and stored. FIG.
In the example of (a) in which the edge pixel 3 with the cross hatching is detected as the starting point, the starting point of the edge is detected by the following method examples (A) to (D) as in the first embodiment. (A) The starting point is searched from the pixel in the upper left corner of the screen. (B) Search for the starting point of the edge along the right lateral direction. (C) If the start point of the edge cannot be found in (B) above, the start point of the edge is searched along the right lateral direction from the left end one pix lower. (D) The above (C) is repeated up to the lower edge of the screen until the start point of the edge can be found.

Next, in step S33 of FIG.
As shown in (b), with the starting point 3 of the edge as the search center,
A range of 1 pix around it is searched clockwise 4 or counterclockwise, and an adjacent edge pixel following the edge start point 3 is searched based on an appropriate threshold value. In FIG. 3 (b), the pixel 5 with cross hatching is the first pixel that satisfies the threshold value,
This is found as an edge pixel and saved.

After the edge pixel 5 is found, the processing proceeds from step S34 to steps S35 and S33 in FIG. 12, where the search center is shifted to the edge pixel 5 and then the surrounding area as shown in FIG. 3 (c). The range of 1 pix is searched clockwise 4 or counterclockwise, and the edge pixel 5 is searched based on an appropriate threshold value.
The next edge pixel following is searched.

By repeating the above-described edge pixel search and search center shift, as shown in FIG. 4, up to the edge pixel 6 with cross hatching can be found and saved.

However, in the example of FIG. 4, when the search center is shifted to the edge pixel 6, the next edge pixel cannot be found even if the range of 1 pix around it is searched.
Conventionally, the edge trace is finished at this stage.

As shown in FIG. 13, with the edge pixel 6 as the center of search, the search range is expanded to a portion outside 1 pix, and the edge pixel is searched by rotating clockwise 4 or counterclockwise within a range of surrounding 2 pix. To do.

That is, when the next edge pixel cannot be found in step S33 of FIG. 12, the search range is expanded to 1 pix outside in step S36, and step S37 described later is executed.
Then, the process proceeds to step S38 via and the edge pixel is searched based on an appropriate threshold value in the search range extended to the surrounding 2pix.

In the example of FIG. 13, edge pixels cannot be found even if the search range is expanded to the outside of 1 pix. Therefore, the process proceeds to step S36 via step S39 in FIG. 12, and as shown in FIG. 14, the search range is further extended to the portion 1 pix outside with the edge pixel 6 as the search center.
Then, in step S38, the next edge pixel is searched in the search range expanded to the surrounding 3 pix.

In the example of FIG. 14, since the edge pixel 9 with the cross hatching can be found, this data is saved and the process proceeds from step S39 to step S40 in FIG. The edge pixel 6 at the search center is interpolated by 2 (= 3-1) pixels 11 and 12 as shown in FIG. 14, and these are also used as edge data.
In the present embodiment, the position of the edge pixel 6 that is the current search center is (X _i , Y _i ), the position of the found edge pixel 9 is (X _{i + k} , Y _{i + k} ), and If the position of the pixel to be interpolated h-th is (X _{i + h} , Y _{i + h} ),
Interpolation of k-1 pixel data is performed by the following equation (6).

(Equation 6)

In the example of FIG. 15, since k = 3, the position (X _{i + 1} , Y _{i + 1} ) of the cross-hatched pixel 11 to be interpolated _first is expressed by the following equation (7). The position of the pixel 12 (X
_{i + 2} , Y _{i + 2} ) is given by the following equation (8).

X _{i + 1} = X _i + (X _{i + 3} −X _i ) / 3 Y _{i + 1} = Y _i + (Y _{i + 3} −Y _i ) / 3 Equation (7)

X _{i + 2} = X _i +2 (X _{i + 3} −X _i ) / 3 Y _{i + 2} = Y _i +2 (Y _{i + 3} −Y _i ) / 3 Equation (8)

When the edge pixel 9 is found, the process proceeds to step S35 in FIG. 12 to shift the search center to this edge pixel 9, and in step S33, the edge pixel is searched for in the range of 1 pix around.

By repeating the above-described edge pixel search, search range expansion, edge pixel interpolation, and search center shift, a long edge chain 13 as shown in FIG. 15 is finally obtained.

Next, steps S37 and S41 of FIG.
Will be described.

In step S37, it is checked to what extent the search range has been expanded. When the pixel is expanded within the predetermined range of the surrounding m pix, the process proceeds to the next step S38 to search for an edge pixel, but if the edge pixel cannot be found even after expanding to the predetermined range, the process proceeds to step S41. Proceed to finish the edge trace. It should be noted that m, which determines the predetermined range, may be changed in a range of 2 or more, and may be set to an appropriate value such as 3 by looking at the edge trace result.

<Fourth Embodiment: Application of Folding Type Edge Tracing Method to Third Embodiment> Next, FIGS.
A fourth embodiment of the present invention will be described with reference to FIGS. 2 to 4 and 12 to 16. The fourth embodiment is an application of the folding type edge tracing method to the third embodiment, and the processing procedure of FIG. 17 is step S4 as compared with that of FIG.
It is the same except that step S55 is added from step 2.
The processing shown in FIG. 17 is also automatically performed by the image processing apparatus. Further, from the image data input in step S30 of FIG. 17, in the next step S31, the edge width formed by the pixel 2 with the diagonal line to the left as shown in FIG. 2 is 1 pixe.
The edge image data 1 of l is created.

First, in steps S32 to S40 of FIG. 17, the same processing as in the third embodiment is performed. The following is a brief summary. In step S32, an edge pixel which is the starting point 3 of the edge trace start is detected based on an appropriate threshold value as shown in FIG. Step S
In 33, the edge pixel 3 is used as the search center and the surrounding 1 pix
The next edge pixel is searched in the range. If found, the process proceeds from step S34 to step S35 to shift the search center, and returns to step S33 to repeat the process of searching for the next edge pixel. In the examples of FIGS. 3 and 4, the edge pixels 5 and 6 can be found one after another. If the next edge pixel cannot be found in the surrounding 1 pix, the process proceeds to step S36, the search range is expanded to the surrounding 2 pix with the edge pixel 6 as the search center as shown in FIG. 13, and the process proceeds to step S37 and step S38 in step S38. The edge pixel is searched, and if it is still not found, the process returns to step S36 via step S39, and the process of further expanding the search range to the surrounding 3 pix as shown in FIG. 14 is repeated. The expansion of the search range is permitted up to the surrounding m pix, and in step S37, it is checked to what extent the search range has been expanded. In the examples of FIGS. 13 to 14, the edge pixels 9 are found in the range of 3 pix around.
Therefore, in step S40, the above equations (6), (7),
Interpolation processing using (8) is performed to interpolate between the found edge pixels 6 and 9 with pixels 11 and 12 as shown in FIG. By such processing, as described above, in the third embodiment, as shown in FIG.
As shown in, the edge chain 13 from the start point 3 to the end point 21 is obtained.

That is, in the third embodiment, step S37.
If it is determined that the search range exceeds the predetermined range m pix, the edge trace is ended.

On the other hand, in the fourth embodiment, the process proceeds from step S37 to step S42 and thereafter, and the edge tracing is further continued. Details are as follows.

That is, when it is determined in step S37 that the search range has exceeded the predetermined range m pix, in step S42 the search center is shifted to the starting point 3 of the edge chain 13 obtained thus far, and in step S43 the search center is shifted. Around 1
An edge pixel of another edge chain existing in the range of pix is searched based on an appropriate threshold value. In step S44, it is determined whether the next edge pixel has been found. If found, the process proceeds to step S45. If not found, the process proceeds to step S45.
Proceed to 54.

In the example of FIG. 16, since an edge pixel of another edge chain 22 cannot be found in 1 pix around the starting point 3, the process proceeds to step S53, and the search range is expanded to 2 pix around the starting point 3 as shown in FIG. 18 (a). , And returns to step S44 through step S55. In the example of FIG.
Edge pixel 2 of another edge chain 22 in the range of pix
I have discovered 3. Therefore, after that, step S45
Proceed to.

The extension of the search range in step S54 is shown in FIG.
2 or similarly to step S36 of FIG. 17, up to the surrounding m pix is permitted, and if the expansion of the search range is checked in step S55 and an edge pixel of another edge chain cannot be found even when it reaches the surrounding m pix, the step Proceed to S41, and end the edge trace.

In the example of FIG. 18A, if m = 3, another edge pixel 23 of the edge chain 22 can be found within the range of 2 pix around, so that the process proceeds from step S44 to step S45, until now. The obtained edge chain 13
Is turned over, the starting point 3 up to this point is set as a new starting point 3A, and the starting point 21 so far is set as a new starting point 21A.

Then, the process proceeds to the next step S46, and the new end point 3A is set as the search center, and the search for the edge pixel subsequent thereto is restarted. However, in the example of FIG.
Since the edge pixel 23 is found in the range extended to pix, the process proceeds to step S53 through steps S49 to 52, and the two edge pixels 3 and 23 are interpolated by the pixel 11 as shown in FIG. 18B.

Next, in step S48, the edge pixel 23
The search center is shifted to, and the process returns to step S46 to search for an edge pixel existing in the range of 1 pix around the new search center based on an appropriate threshold value. By repeating this process, in the example of FIG. 18, as shown in (b) of FIG.
Is even discovered.

After shifting the search center to the edge pixel 24, the next edge pixel cannot be found in the surrounding 1 pix, so the process proceeds from step S47 to step S49, and the search range is changed to 1 pix up to the surrounding 2 pix without changing the search center. It is expanded by an amount, and the search is performed based on an appropriate threshold value through step S50. If no edge pixel can be found in step S52, the process returns to step S49 to further expand the search range and continue the search. If the next edge pixel can be found by the expansion, the pixels in between are interpolated in step S53 as in step S40 of FIG. 12 or FIG.
The edge pixel in which the search center is found is shifted in step 48, and the process returns to step S46.

The expansion of the search range in step S49 is permitted up to the surrounding m pix. If the expansion of the search range is checked in step S50 and the next pixel cannot be found even when the surrounding m pix is reached, the process proceeds to step S41. Advance Ends edge trace.

With the above processing, in the example of FIG. 18B, the edge pixel 24 becomes the end point. That is, the starting point 21A
A long edge chain 25 from the end point 24 to the end point 24 is obtained.

In each of the first to fourth embodiments described above, it is assumed that the specific edge of the object is detected from the edge image data by image processing. However, the present invention can also be applied to the case of detecting an abstract edge. That is, since the edge pixel may be any pixel as long as it represents a certain limit, a pixel satisfying a certain specific condition in an arbitrary spatial image data group is an edge pixel.
A series of such edge pixels is also an edge chain, and the present invention can be applied.

For example, it is assumed that water temperature data regarding a three-dimensional position in a pool or a lake is obtained. In this case, considering a case where a chain of 10 to 12 ° C. on the water surface is traced as an edge chain, the temperature distribution on the water surface is given as image data.
An edge chain satisfying the condition of 10 to 12 ° C. can be obtained as in the first to fourth embodiments only by using pixels of 0 to 12 ° C. as edge pixels. Further, if it is expanded three-dimensionally, a specified temperature distribution in water can be detected as an edge chain.

[0086]

According to the first aspect of the present invention, when the edge pixel cannot be found, the position where the next edge pixel is likely to exist is estimated from the already obtained edge pixel position data, and the search center is set to the pixel at the estimated position. Since the next edge pixel is searched for as a shift, a relatively long edge chain can be obtained. As the estimation method, the method based on the linear approximation as in the second invention may be simple, and for example, the formula (1) in the third invention may be used. In the fourth invention, when the edge pixel cannot be found even if the estimation is continuously performed a predetermined number of times, the search for the edge pixel is ended, so that the edge image data is subdivided by noise and It can be distinguished from the end. In the fifth aspect of the invention, since the final position of the edge chain is returned to the edge pixel that was found last, the reliability of the edge chain is high.

On the other hand, in the sixth invention, when the edge pixel cannot be found, the search range is kept 1 pi without changing the search center.
Edge pixels are sequentially searched from the range of x by 1 pix to a range of surrounding m pix in a corrugated form, and when an edge pixel is found in the search range of surrounding k pix, the middle is interpolated with k-1 pixel data. Therefore, a relatively long edge chain can be obtained. As for the interpolation method, as in the seventh invention, the interpolation by the linear approximation using the equation (2) is easy.

Further, in the eighth to tenth inventions, even if the edge trace is started in the middle of the edge chain, a long edge chain can be obtained without being divided at the starting point.

[Brief description of drawings]

FIG. 1 is a flowchart showing an edge tracing method according to a first embodiment.

FIG. 2 is a diagram showing an example of edge image data.

FIG. 3 is a diagram showing a state in which a next edge pixel is found within a range of 1 pix around the pixel.

FIG. 4 is a diagram showing a state in which a next edge pixel cannot be found in a surrounding 1 pix range.

FIG. 5 is a diagram showing edge pixel positions estimated for the first time.

FIG. 6 is a diagram showing how the search center is shifted to the edge pixel estimated for the first time and the next edge pixel is searched.

FIG. 7 is a diagram showing edge pixel positions estimated for the second time.

FIG. 8 is a diagram showing how the search center is shifted to the edge pixel estimated for the second time and the next edge pixel is searched.

FIG. 9 is a view showing the obtained edge chain.

FIG. 10 is a flowchart showing an edge tracing method according to the second embodiment.

FIG. 11 is a diagram showing how the edge chain is folded back to search for the next edge pixel.

FIG. 12 is a flowchart showing an edge tracing method according to the third embodiment.

FIG. 13 is a diagram showing how the search range is expanded to the outside of 1 pix to search for the next edge pixel.

FIG. 14 is a diagram showing how the search range is further expanded to the outside of 1 pix to search for the next edge pixel.

FIG. 15 is a diagram showing interpolated edge pixels.

FIG. 16 is a view showing the obtained edge chain.

FIG. 17 is a flowchart showing an edge tracing method according to the fourth embodiment.

FIG. 18 is a diagram showing how an edge chain is folded back to search for the next edge pixel.

FIG. 19 is a flowchart showing a conventional method.

FIG. 20 is a diagram showing edge chain division in the conventional technique.

FIG. 21 is a diagram showing edge chain division in the related art.

FIG. 22 is a diagram showing the principle of a folded edge tracing method according to the present invention.

FIG. 23 is a diagram showing the principle of a folded edge tracing method according to the present invention.

FIG. 24 is a diagram showing the principle of a folded edge tracing method according to the present invention.

FIG. 25 is a diagram showing the principle of a folded edge tracing method according to the present invention.

[Explanation of symbols]

 1 Edge Image Data 3 Edge Start Point 3A New End Point 5,6,9,21,23,24 Discovered Edge Pixel 7,8 Estimated Edge Pixel 10,13,20,25 Edge Chain 11,12 Interpolated Edge Pixel 21A New starting point 22 Another edge chain

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Koji Ikura 2-17-1 Osaki, Shinagawa-ku, Tokyo Inside the Meidensha Co., Ltd.

Claims (10)

[Claims] 1. A process for searching a next edge pixel existing in a range of 1 pix around the pixel satisfying a specific condition as an edge pixel and using the edge pixel as a search center by sequentially shifting the search center. By repeating
In an edge tracing method for obtaining an edge chain: if the next edge pixel can be found in the search for the edge pixel, shift the search center to this found edge pixel; If the edge pixel of is not found, the position of the next edge pixel is estimated from the position data of the edge pixel already obtained, and the search center is shifted to the pixel at this estimated position. 2. The edge tracing method according to claim 1, wherein the next edge pixel position is estimated by linear approximation. 3. The method according to claim 1, wherein the position of the current edge pixel is (X _i , Y _i ), and the position of the jth previous edge pixel is (X _ij , Y _ij ), and An edge tracing method characterized in that, when the position is (X _{i + 1} , Y _{i + 1} ), it is estimated from the position data of the past n edge pixels by the following expression (1). [Equation 1] 4. The edge pixel search according to claim 1, wherein the edge pixel search is terminated when the edge pixel position cannot be found from the image data even if the next edge pixel position estimation is continued a predetermined number of times. An edge tracing method characterized by the above. 5. The edge trace method according to claim 4, wherein when the search for the edge pixel is completed, the final position of the edge chain is returned to the edge pixel that was found last. 6. A process of searching a next edge pixel from image data in a search range around the edge pixel, which is a pixel satisfying a specific condition, is set as a search center, and the found edge pixel is sequentially searched as a search center. In the edge tracing method in which an edge chain is obtained by shifting and repeating: the search range at the same search center is 1pix from the range of 1pix around until the next edge pixel can be found.
When m is an integer of 2 or more and k is an integer of m or less, when an edge pixel can be found in the search range of the surrounding k pix, the found edge pixel and the edge pixel at the search center And k−1 pixels are interpolated between the distances between and, and when the edge pixel cannot be found in the search range of the surrounding m pix, the edge pixel search is ended. 7. The position of the edge pixel which is the current search center is (X _i , Y _i ) and the position of the found edge pixel is (X _{i + k} , Y _{i + k} ) according to claim 6. Assuming that the position of the pixel to be interpolated h-th from the current search center is (X _{i + h} , Y _{i + h} ), k−1 is calculated by the following equation (2).
An edge tracing method, characterized by performing interpolation of individual pixels. [Equation 2] 8. A process in which a pixel satisfying a specific condition is set as an edge pixel, the edge pixel is set as a search center, and the next edge pixel existing in a specific range around the edge pixel is searched from the image data by sequentially shifting the search center. In the edge tracing method that obtains an edge chain from the start point to the end point by repeating: If an edge pixel at the end point is found, is there an edge pixel of another edge chain near the start point of the edge chain obtained so far? Determining whether or not;
If the edge pixel of the another edge chain exists, the start point and the end point of the edge chain obtained so far are replaced, and the search for the edge pixel is continued from the new end point; Terminating the search for edge pixels if they do not exist; 9. The method according to claim 1, wherein
If the edge pixel position cannot be found from the image data even if the estimation of the next edge pixel position continues for a predetermined number of times, the edge pixel of another edge chain exists near the start point of the edge chain obtained up to now. If there is an edge pixel of the other edge chain, the start point and the end point of the edge chain obtained so far are replaced, and the search for the edge pixel is continued from the new end point; Terminating the search for an edge pixel when there is no edge pixel in the other edge chain; 10. The edge pixel of another edge chain according to claim 6 or 7, when an edge pixel cannot be found in the search range of the surrounding m pix, in the vicinity of the start point of the edge chain obtained so far. If there is an edge pixel of the other edge chain, the start point and the end point of the edge chain obtained up to now are exchanged, and the search for the edge pixel is continued from the new end point. An edge tracing method, wherein the edge pixel search is terminated when there is no edge pixel of the another edge chain.