JPH07129767A - Line segment sampling method - Google Patents

Abstract

PURPOSE:To sample a line segment with high accuracy and to shorten the processing time by generating a histogram, for a point where the endpoint of an image interscets with all straight lines connecting two points in the image, and sampling the straight line connecting a point which takes the maximum value of the histogram to a point which takes a second maximum value as the line segment. CONSTITUTION:An edge candidate point image shown with a black point can be acquired as the process result of an edge candidate point sampling part. A line segment sampling part finds the point where the straight line connecting the two points for all combination of the two points of the edge candidate point image crosses the endpoint, and generates the histogram representing the number of times of appearance of an intersection. For example, the histogram 2 of the intersections C1, C2 of the straight line 1 passing the points p1, p2 and the endpoint is increased. An intersection histogram can be acquired by repeating such processing on the combination of all edge candidate points. The straight line L1 connecting the point (a) taking the maximum value to the point (b) taking the second maximum value is sampled as the line segment, then, the endpoint coordinate o the line segment is outputted.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique for extracting a line segment in an image from information on characteristic points in the image.

[0002]

2. Description of the Related Art When extracting a line segment from information on a feature point in an image, generally, the Hough transform or the least square approximation method is used. However, such a method has problems that the processing is slow, a huge memory is required for the processing, or the extraction accuracy is insufficient.

In view of such a problem, there is a method (hereinafter, method A) for extracting a line segment in an image from a histogram of an intercept of a straight line connecting two feature points in the image with respect to X and Y coordinate axes. Proposed (BEHZAD K
AMAGAR-PARSI, BEHROOZ KAMA
GAR-PARSI, and NATHAN S. NETA
NYAHU, "A nonparametric Method for Fittin
ga Straight Line toa Noisy Image ", IEEE
TRANSACTION ON PATTERNAL NAL
YSIS AND MACHINE INTELLIGEN
CE, Vol. 11, no.9, SEPTEMBER 198
9).

[0004]

The method A described in the above document is advantageous in that the line segment can be extracted with a relatively short processing time and a small memory as compared with the conventional general method. have. However, the reference axes (X, Y
There is a problem that a line segment parallel to the coordinate axis) cannot be extracted, and a line segment with a tilt close to the reference axis has a large extraction error. Further, the above-mentioned document describes only the case of extracting one line segment from the image, and does not refer to the method of extracting the plurality of line segments from the image. It cannot be applied immediately to the purpose.

An object of the present invention is to provide a method capable of extracting a line segment in an image with high accuracy and a plurality of line segments sequentially extracted from the image with a relatively short processing time and a small memory. To do.

[0006]

According to the line segment extracting method of the present invention, a histogram of points intersecting the end points of the image is created for all straight lines connecting two points in the image, and the maximum value of the histogram is generated. It is characterized in that a straight line that connects the point that takes a point with the point that takes the second largest value is extracted as a line segment.

According to the line segment extraction method of the present invention, a histogram of points intersecting the end points of the image is created for all straight lines connecting two points in the polygonal image surrounded by a plurality of sides, It is characterized in that a straight line connecting a point having the maximum value of the created histogram and a point having the maximum value of the histogram on the side not including the point is extracted as a line segment.

According to the line segment extraction method of the present invention, a histogram α of points intersecting the end points of the image is created for all straight lines connecting two points in the image, and this histogram α
Create a histogram β of points that intersect the end points of the image for all straight lines passing through the maximum value of, and connect the straight line connecting the maximum value of this histogram β and the point with the second largest value. It is characterized in that it is extracted as a minute.

According to the line segment extraction method of the present invention, a histogram α of points intersecting the end points of the image is created for all straight lines connecting two points in a polygonal image surrounded by a plurality of sides. , The histogram β of the points that intersect with the end points of the image of all the straight lines that pass through the point of maximum value of this histogram α
Is generated for an edge that does not include the point, and the histogram α
It is characterized in that a straight line connecting the point having the maximum value of and the point having the maximum value of the histogram β is extracted as a line segment.

The invention according to claim 5 is the invention according to claim 3 or 4.
In the line segment extraction method of the invention described in the paragraph, the histogram corresponding to the extracted line segment is deleted from the histogram α created when one line segment is extracted, and the next line segment is extracted using the remaining histogram. It is characterized in that it is extracted by the same method as the line segment extracted previously.

The invention according to claim 6 is the invention according to claim 3 or 4.
In the line segment extraction method of the invention described, the histogram β is subtracted from the histogram α created when one line segment is extracted, the remaining histogram is used as the histogram α, and the next line segment is extracted before. It is characterized by being extracted by a method similar to.

According to a seventh aspect of the invention, in the line segment extracting method according to the fifth aspect of the invention, the histogram corresponding to the line segment extracted from the histogram α is deleted, and the maximum value of the remaining histogram is a predetermined threshold value. The feature is that the line segment extraction process is terminated when the size becomes smaller.

According to an eighth aspect of the invention, in the line segment extracting method according to the sixth aspect of the invention, when the maximum value of the histogram remaining after subtracting the histogram β from the histogram α becomes smaller than a predetermined threshold value, The line segment extraction process is terminated.

According to a ninth aspect of the present invention, in the line segment extracting method according to the fourth aspect, for a side including a point having the maximum value of the histogram α, a histogram corresponding to the extracted line segment is used as a histogram. After deleting from α, the remaining histogram is used again as the histogram α, and for the side that does not include the point having the maximum value of the histogram α, the histogram β is subtracted from the histogram α and the remaining histogram is used as the histogram α again.
The feature is that the next line segment is extracted by the same method as the line segment extracted previously.

The invention according to a tenth aspect is the first to ninth aspects.
In the line segment extraction method of the invention described in any one of
It is characterized in that a histogram of the coordinates of points in the image corresponding to the extracted line segment is created, and the end points of the line segment are detected from this histogram.

[0016]

According to the method of the first aspect of the present invention, the histogram of the points where the straight line connecting all two points in the image and the end point (frame) of the image intersect is created. A straight line connecting the point with the maximum value of this histogram and the point with the next largest value is extracted as a line segment, and this line segment is generally regarded as the line segment that best represents the set of points in the image. be able to.

Unlike the method A, since the reference axis is not used, it becomes impossible to extract a line segment in a specific direction.
Alternatively, it is possible to avoid the disadvantage that the extraction accuracy is extremely deteriorated.

Here, the histogram will be described. Two
When extracting a line segment by paying attention to the intersection between the straight line connecting the points and the end point of the image, the most common method is that one straight line connecting the two points intersects the end point of the image at two points. So, as a pair of two intersections about one straight line,
It may be a method of creating a histogram of the appearance frequency of each pair, detecting the pair of intersections having the maximum value, and extracting the straight line connecting the two points as a line segment. This method using a two-dimensional histogram has an advantage in extracting a plurality of line segments, as described later, but it requires a very large memory for creating the histogram. Since the access time also increases, the processing time becomes slow.

On the other hand, the histogram created by the method according to the first aspect of the present invention is a one-dimensional histogram representing the number of appearances at each point (individual point) where the straight line connecting the two points and the end point of the image intersect. Therefore, there is an advantage that the size of the memory required for creating the histogram is much smaller and the access time is also reduced, so that the processing time is also faster.

According to the method of the invention of claim 2, claim 1
Similarly to the invention of (1), a one-dimensional histogram of the appearance frequency of each point intersecting the straight line connecting the two points and the end point of the image is created, and the line segment is extracted based on this histogram. However, in the present invention, since an image surrounded by a polygonal frame having a plurality of sides is targeted, a line segment to be extracted should pass through points on two different sides. . Therefore, the point that takes the maximum value of the created histogram,
A straight line that connects the point having the maximum value in the histogram on the side not including this point is extracted as a line segment that best represents the set of points in the image.

According to the method of the invention of claim 3, claim 1
Similar to the method of the invention described above, a histogram α of the appearance frequency of the points where the straight line connecting the two points in the image and the end points of the image intersect is created.

Here, since the histogram α is a one-dimensional histogram representing the frequency of appearance of each intersecting point,
There is no guarantee that the point that takes the maximum value and the point that takes the next largest value are the pair of points where the straight line connecting the same two points and the end point of the image intersect (this is more than the two-dimensional histogram above). This is a disadvantage). If the first line segment that most represents the set of points in the image is limited, the probability that the two points form a pair is sufficiently high. Therefore, when extracting only one line segment, it is practical. There is no problem even if the line segment is extracted using only the one-dimensional histogram. However, when two or more line segments are sequentially extracted, the risk of erroneously extracting a straight line connecting two unpaired points is increased for the second and subsequent line segments.

In order to solve such a problem, according to the method of the present invention, among the straight lines connecting two points in the image,
The histogram β of the appearance frequency of the points intersecting with the end points of the image is created by limiting to the straight line passing through the point having the maximum value of the histogram α. Then, a straight line connecting a point passing through the maximum value of the histogram β (which coincides with the maximum value of the histogram α) and a point having the next largest value is extracted as a line segment. Since the histogram β is related to a straight line that always passes through the maximum value, the point that takes the second largest value is a point that always pairs with the point that takes the maximum value.
Therefore, the above problem is avoided.

According to the invention of claim 4, as in the method of the invention of claim 2, one-dimensional representing the appearance frequency of each point where the straight line connecting two points in the image and the end point of the image intersect. A histogram α of is created. Next, of the straight lines connecting the two points in the image, the one-dimensional histogram β representing the appearance frequency of the points intersecting the end points of the image is limited to the straight line passing through the point having the maximum value of the histogram α. But the histogram α
Created for edges that do not contain points that take the maximum of. Then, a straight line connecting the point having the maximum value of the histogram α and the point having the maximum value of the histogram β is extracted as a line segment.

As can be understood from the description related to the invention of claim 3, there is no guarantee that the point having the maximum value of the one-dimensional histogram α and the point having the next largest value form a pair. . On the other hand, it is guaranteed that the point having the maximum value of the histogram α and the point having the maximum value of the histogram β are paired points. Therefore, according to the method of the present invention, like the method of the third aspect of the present invention, even when two or more line segments are sequentially extracted, the line segment can be reliably extracted.

According to the method of the inventions of claims 5 to 10,
Two or more line segments representing a set of points in the image can be sequentially extracted.

In the method of the invention of claim 5 or 6, basically, the line segments are sequentially extracted by the method of the invention of claim 3 or 4, but in the invention of claim 5, the previous line is extracted. The histogram corresponding to the extracted line segment is deleted from the histogram α created during the segment extraction, and the remaining histogram is used as the histogram α for extracting the next line segment. In that case, the histogram remaining after subtracting the histogram β from the histogram α created at the time of extracting the previous line segment is used as the histogram α for extracting the next line segment.

In the method of the ninth aspect of the invention, basically, the line segments are sequentially extracted by the method of the fourth aspect of the invention, but the histogram α created in the previous line segment extraction is used. For the side that includes the point with the maximum value, the histogram corresponding to the extracted line segment is deleted from the histogram α, and the remaining histogram is used as the histogram α for the extraction of the next line segment. For a side that does not include a value-taking point, the histogram β is subtracted from the histogram α and the remaining histogram is used as the histogram α for extracting the next line segment.

As described above, according to the method of the present invention as defined in claim 5, 6 or 9, the histogram α or β for extracting the previous line segment is used to extract the histogram α for extracting the next line segment.
Is generated by a simple calculation process, the processing time for creating the histogram can be reduced and the efficiency of the line segment extraction process can be improved.

According to the method of the invention of claim 7, claim 5
In the method of the invention, the maximum value of the histogram remaining after deleting the histogram corresponding to the line segment extracted from the histogram α is compared with a certain threshold value, and when the line segment becomes smaller than this threshold value, Abort the extraction process. Further, according to the method of the invention of claim 8, in the method of the invention of claim 6, the histogram remaining after subtracting the histogram β from the histogram α is compared with a certain threshold value and becomes smaller than this threshold value. At this stage, the line segment extraction process is terminated. In this way, the end of processing can be controlled so that only the line segment to be extracted is extracted.

The end points of the line segment extracted by the method according to the first to ninth aspects coincide with the end points of the image, but this is not necessarily the original end point of the line segment. According to the method of the invention of claim 10, a histogram of the coordinates of the points in the image corresponding to the line segment extracted by the method of the invention of any one of claims 1 to 9 is created, and this histogram twist line is drawn. The end points of the minutes can be detected more accurately.

[0032]

EXAMPLES Examples of the present invention will be described below. FIG. 1 shows an example of a system configuration for each embodiment described below.

In FIG. 1, a color image input apparatus (color scanner) 100 reads a color image original and inputs the image data as RGB data.
The brightness calculation unit 102 calculates the brightness L of each pixel from the input RGB data by the following equation, and outputs the brightness image data of the input image.

L = (R + G + B) / 3 The edge candidate point extraction unit 104 extracts a point (edge candidate point) having a large luminance slope as a feature point in the luminance image. In each of the following embodiments, the quadratic difference mask shown in FIG. 2 is used to calculate the quadratic difference value of each pixel on the luminance image, and the point having the quadratic difference value of 0 is extracted as an edge candidate point. .

For example, in the area where the brightness of each pixel is as shown in FIG. 3, when the secondary difference mask of FIG. 2 is applied to the central pixel, the secondary difference value S is S = 4 * 72- (76 + 76 + 68 + 68). = 0. Therefore, this center point is extracted as an edge candidate point.

When the brightness of each pixel is in the area shown in FIG. 4, the secondary difference value S of the central pixel is S = 4 * 72- (76 + 88 + 71 + 68) = 16, which is not zero. Therefore, this center point is not an edge candidate point.

In this way, the edge candidate point extraction unit 10
In step 4, the edge candidate points in the luminance image are extracted, and the edge candidate point image (white image except for the edge points) is sent to the line segment extracting unit 106. The line segment extraction unit 106 extracts a line segment by integrating the edge candidate points in the edge candidate point image using the algorithm of the present invention, and outputs line segment data (coordinates at both ends of the line segment). The contents of this line segment extraction processing will be described below for each embodiment.

<Embodiment 1> Here, it is assumed that an edge candidate point image as shown in FIG. 5 is obtained as a processing result of the edge candidate point extracting unit 104. In FIG. 5, black points mean edge candidate points. Further, in the example of this image, there is a circular frame (end point of the image), but the shape of the frame is basically arbitrary.

In the line segment extracting unit 106, first, for all combinations of two points in the edge candidate point image, a point where the straight line connecting the two points and the end point of the edge candidate point image intersect is obtained, and the intersection point Create a histogram of the number of occurrences of. FIG. 6 is an explanatory diagram thereof.

As shown in FIG. 6A, for example, points p1,
The histograms of intersections c1 and c2 between the straight line L1 passing through p2 and the end points of the edge candidate point image (FIG. 5) are incremented. By repeating this processing for all combinations of edge candidate points, the intersection histogram shown in FIG. 6B is obtained.

In the intersection histogram thus obtained, a straight line connecting the point a having the maximum value and the point b having the second largest value, that is, the straight line L in FIG. 6A.
1 is extracted as a line segment, and the end point coordinates and the like are output.

<Embodiment 2> Here, it is assumed that the edge candidate point image shown in FIG. 7 is input to the line segment extracting unit 106. In FIG. 7, black points mean edge candidate points. Although the example of this image has a square frame (end points of the image), an image having an arbitrary polygonal frame may be processed in the same manner.

In the line segment extracting unit 106, first, for all combinations of two points on the edge candidate point image, a point at which the straight line connecting the two points and the end point of the edge candidate point image intersect is obtained, and the intersection point A histogram of the number of appearances is created for each side A, B, C, D of the edge candidate point image.

Next, a line segment connecting the point having the maximum value of the intersection histogram and the point having the maximum value of the histogram is extracted from the sides other than the side where the point exists, and the coordinates of the end points are extracted. Output.

FIG. 8 is an explanatory diagram of this line segment extraction. It is assumed that an intersection histogram as shown in FIG. 8B is obtained for the edge candidate point image of FIG. The point having the maximum value in the intersection histogram is the point a on the side A. Among the sides B, C, and D where the point a does not exist, the point where the histogram has the maximum value is the point c on the side B. Therefore, in this example, the straight line connecting the points a and c, that is, the straight line L1 in FIG. 8A is extracted as the line segment.

<Third Embodiment> Here, it is assumed that the edge candidate point image shown in FIG. 7 is input to the line segment extracting unit 106.

In the line segment extracting unit 106, first, for all combinations of two points on the edge candidate point image, the straight line connecting the two points and the end point of the edge candidate point image, as in the second embodiment. Intersection points are obtained, and a histogram of the number of appearances of intersections (referred to as intersection histogram α) is created for each side A, B, C, D of the edge candidate point image. Here, the intersection histogram α shown in FIG.
Is obtained.

Next, of the straight lines connecting the two points in the edge candidate point image, the point having the maximum value of the intersection histogram α, here, the straight line passing through the point a, of the points intersecting the end points of the edge candidate point image. A histogram of the number of appearances (called an intersection histogram β) is created. FIG. 10 is an explanatory diagram thereof.

In FIG. 10A, since the straight line L1 passing through the points p1 and p2 passes through the point a, the histograms of the points a and e which are the intersections with the end points of the edge candidate point image are incremented. On the other hand, since the straight line L2 passing through the points p3 and p4 does not pass through the point a, the histogram of the intersection with the end point of the edge candidate point image is not updated. By repeating the same operation, here, FIG.
It is assumed that the intersection histogram β shown in (B) is obtained.

Finally, the point at which the maximum value of the intersection histogram β obtained as described above is taken, and 2 on the side different from this point.
A straight line connecting the point with the second largest value, here a straight line connecting the points a and c, is extracted as a line segment,
Data such as the end point coordinates are output.

<Embodiment 4> Here, it is assumed that the edge candidate point image shown in FIG. 7 is input to the line segment extracting unit 106.

In the line segment extraction unit 106, first, for all combinations of two points on the edge candidate point image, the straight line connecting the two points and the end points of the edge candidate point image, as in the second embodiment. Intersection points are obtained, and a histogram of the number of appearances of intersections (referred to as intersection histogram α) is created for each side A, B, C, D of the edge candidate point image. Here, the intersection histogram α shown in FIG.
Is obtained.

Next, among the straight lines connecting the two points in the edge candidate point image, the point having the maximum value of the intersection histogram α, in this case, the straight line passing through the point a, the sides other than the side including the point a ( Here, a histogram (referred to as an intersection histogram β) of the number of appearances of points intersecting the sides B, C, and D) is created. FIG. 11 is an explanatory diagram thereof.

In FIG. 11A, since the straight line L1 passing through the points p1 and p2 passes through the point a, the histogram of the point e which is the intersection with the end point of the edge candidate point image is incremented. However, since the point a is a point on the side A, the histogram is not created. On the other hand, point p
Since the straight line L2 passing through 3 and p4 does not pass through the point a,
Do not update the histogram. It is assumed that the intersection histogram β shown in FIG. 10B is obtained by repeating the same operation.

Finally, the intersection histogram α (see FIG. 9)
(A) which takes the maximum value in (B)) and the intersection histogram β
The line segment connecting the point having the maximum value of, the point c'here, is extracted, and the data of the end point coordinates and the like is output.

<Fifth Embodiment> Here, it is assumed that the edge candidate point image shown in FIG. 7 is input to the line segment extracting unit 106.

In the line segment extraction unit 106, first, the same process as in the third embodiment is performed, and the first line segment, here, a
A line segment (1) connecting the point and the point c is extracted. A description of this extraction processing is omitted.

The portion corresponding to the line segment (1) is deleted from the intersection histogram α obtained when the first line segment (1) is extracted. FIG. 12 is an explanatory diagram thereof.

That is, in this example, the intersection histogram α shown in FIG. 12 (A) is obtained. From this histogram, points that take local minimum values on both sides from point a, which is one end point of the first line segment. , Here, the points e and f are obtained, and the histogram between these points is deleted. Similarly, from the point c, which is the other end point of the first line segment (1), the minimum points on both sides,
Here, g point and h point are obtained, and the histogram between both points is deleted.

By correcting the intersection histogram α in this manner, the intersection histogram γ shown in FIG. 12B is obtained.

Next, the intersection histogram γ obtained by correcting the intersection histogram α is used again as the intersection histogram α, and the second line segment is extracted by the same process as the first line segment. Here, of the straight lines connecting the two points in the edge candidate point image, a histogram of the number of appearances of intersections between the straight line passing through the point b having the maximum value of the intersection histogram γ and the end point of the edge candidate point image, that is, FIG. A histogram corresponding to the intersection histogram β shown in (B) is created, and a line segment that connects the point having the maximum value and the point having the second largest value on the side other than the side where this point exists Is extracted as the second line segment (2). In this example, the line segment connecting the points b and d shown in FIG. 12B would be extracted as the second line segment.

<Sixth Embodiment> Here, it is assumed that the edge candidate point image shown in FIG. 7 is input to the line segment extracting unit 106.

In the line segment extracting unit 106, first, the same process as in the third embodiment is performed to extract the first line segment, here the line segment (1) connecting the points a and c. A description of this extraction processing is omitted.

Next, the second line segment (2) is extracted. FIG. 13 is an explanatory diagram thereof. FIGS. 13A and 13B show the intersection histogram α and the intersection histogram β obtained when the first line segment (1) is extracted.

First, a histogram (referred to as an intersection histogram δ) obtained by subtracting the intersection histogram β from the intersection histogram α is obtained. Here, the intersection histogram δ shown in FIG. 13C is obtained.

Then, this intersection histogram δ is used again as an intersection histogram α, and the second line segment is extracted by the same processing as the first line segment. Here, of the straight lines connecting the two points in the edge candidate point image, a histogram of the number of appearances of the intersections between the straight line passing through the point b having the maximum value of the intersection histogram δ and the end point of the edge candidate point image, that is, FIG. A histogram corresponding to the intersection histogram β shown in (B) is created, and a line segment that connects the point having the maximum value and the point having the second largest value on the side other than the side where this point exists Is extracted as the second line segment (2). In this example, the line segment connecting points b and d shown in FIG. 13C would be extracted as the second line segment.

It should be noted that in the present embodiment, the side corresponding to the line segment extracted from the histogram for the side where the point having the maximum value of the intersection histogram α obtained at the time of extracting the first line segment exists. The histogram obtained by subtracting is used as the corrected histogram α (corresponding to the histogram δ in FIG. 13C) for the relevant side, and for the other sides, the remaining histogram obtained by subtracting the histogram β from the histogram α is the corrected histogram α. As (corresponding to the histogram δ in FIG. 13C), the same 2
It is also possible to modify so as to perform the line segment extraction of the real line.

<Embodiment 7> Here, it is assumed that the edge candidate point image shown in FIG. 7 is input to the line segment extracting unit 106.

In the line segment extraction unit 106, first, the same process as in the third embodiment is performed, and the first line segment, here, a
A line segment (1) connecting the point and the point c is extracted. A description of this extraction processing is omitted.

The portion corresponding to the line segment (1) is deleted from the intersection histogram α obtained when the first line segment (1) is extracted. FIG. 12 is an explanatory diagram thereof.

That is, in this example, the intersection histogram α shown in FIG. 12 (A) is obtained. From this histogram, points that take local minimum values on both sides from point a, which is one end point of the first line segment, are obtained. , Here, the points e and f are obtained, and the histogram between these points is deleted. Similarly, from the point c, which is the other end point of the first line segment (1), the minimum points on both sides,
Here, g point and h point are obtained, and the histogram between both points is deleted.

By correcting the intersection histogram α in this manner, the intersection histogram γ shown in FIG. 12B is obtained.

Next, it is checked whether the maximum value m of the intersection histogram γ obtained by modifying the intersection histogram α is larger than a predetermined threshold Th, and if m> Th, the processing is continued.

Here, the intersection histogram γ shown in FIG. 12B is obtained, and its maximum value m is 223 (point b). When Th = 200, m> Th, and therefore the processing is continued.

That is, the intersection histogram γ shown in FIG. 12B is set again as an intersection histogram α, and a point having the maximum value thereof, here, a straight line connecting two points passing through the point b, intersects with the end point of the edge candidate point image. A histogram of the number of appearances of is generated, and the line segment connecting the point having the maximum value and the point having the second largest value is extracted as the second line segment (2). FIG. 14 is an explanatory diagram thereof.

In FIG. 14A, since the straight line L1 passing through the points p1 and p2 passes through the point b, the histograms of the points b and k which are the intersections with the end points of the edge candidate point image are incremented. On the other hand, since the straight line L2 passing through the points p3 and p4 does not pass through the point b, the histogram is not updated. It is assumed that the intersection histogram β shown in FIG. 14B is obtained in this way. In this case, the line segment connecting the point b having the maximum value and the point d having the second largest value is extracted as the second line segment (2).

After extracting the line segment (2) in this way,
Regarding the intersection histogram α at that time, that is, the intersection histogram γ in FIG. 12B, the points between the minimum values on both sides of the end point of the line segment (2) are deleted, and the corrected intersection histogram is obtained. create. FIG. 15 is an explanatory diagram thereof.

FIG. 15A is an intersection histogram α at the stage when the line segment (2) is extracted (this is the same as the intersection histogram γ shown in FIG. 12B). From this intersection histogram, by deleting between the two points having the minimum values on both sides of the end point b of the line segment (2) and the two points having the minimum values on both sides of the other end point d of the line segment (2), , The intersection histogram shown in FIG. 15 (B) is obtained. The maximum value m of this histogram is 20 (point i), which is Th = 200.
Since it is smaller, the line segment extraction process ends.

If m> Th, the intersection histogram of FIG. 15B is used again as the intersection histogram α, and the similar line segment extraction processing is continued. That is, the intersection histogram β is created for a straight line passing through the point having the maximum value of the intersection histogram α, and the line segment connecting the point having the maximum value and the point having the second largest value is defined as the next line. Extract as minutes. Hereinafter, similar processing is continued while m> Th.

<Embodiment 8> In the line segment extraction unit 106, the first
As a step, the embodiment 3 (or the other embodiment described above)
A line segment is extracted by the same processing as. The endpoints of the line segments extracted in this way are all equal to the endpoints of the edge candidate point image. In the present embodiment, the line segment extracting unit 106 performs the process of determining the end points of the line segment extracted in the first stage as the second stage. FIG. 16 is an explanatory diagram thereof.

The line segment L having the end points a and c as shown in FIG. 16A is the line segment extracted in the first stage. In the second stage, a histogram of the coordinates of the edge candidate points on this line segment L is created. Then, the coordinates having a value of 90% of the maximum value of this histogram are detected and used as the end points of the line segment. That is, when the histogram shown in FIG. 16B is obtained, the points j and k are detected as the end points of the line segment L. In other words, a line segment whose end points are the points j and k is finally extracted.

In this embodiment, the histogram of the edge candidate points on the line segment extracted in the first step is used. For example, in the band area where the distance from the extracted line segment is less than a certain threshold value. It is also possible to create a histogram for the edge candidate points (inside the area indicated by the broken line in FIG. 16A) and detect the end points in the same manner from this histogram.

Further, it is also possible to extract the portion corresponding to the line segment extracted from the intersection histogram and use the histogram of the edge candidate points corresponding to the histogram for the end point detection.

[0084]

As described above, according to the present invention, the following effects can be obtained.

According to the first to tenth aspects of the present invention, the line segment representing the set of points in the image can be accurately extracted. Further, since the method does not use the reference axis, it is possible to avoid inconvenience that it becomes impossible to extract a line segment in a specific direction, or the extraction accuracy is extremely deteriorated. Since a two-dimensional histogram representing the frequency of appearance of a point where a straight line connecting two points in an image and an end point of the image appear is used, the memory size for creating the histogram can be reduced and the time required for accessing the histogram can be reduced. Since it can be shortened, the processing time can be shortened.

According to the third or fourth aspect of the invention, the weakness of the method of using the one-dimensional histogram is compensated by creating the histogram in two steps, and the purpose is to sequentially extract two or more line segments. Also, reliable line segment extraction is possible.

According to the methods of the inventions of claims 5 to 10,
Two or more line segments representing a set of points in the image can be sequentially extracted.

According to the fifth, sixth or ninth aspect of the present invention, the histogram α for extracting the next line segment can be created from the histogram α or β for extracting the previous line segment by a simple calculation process. As a result, the processing time for creating the histogram can be reduced and the efficiency of the line segment extraction processing can be improved.

According to the invention of claim 7 or 8, it is possible to control the end of the processing so as to extract only the line segment to be extracted based on the comparison result of the maximum value of the histogram and the threshold value.

According to the tenth aspect of the invention, it is possible to detect a more accurate end point of a line segment representing a set of points in an image.

[Brief description of drawings]

FIG. 1 shows an example of a system configuration for each embodiment.

FIG. 2 shows an example of a quadratic difference mask for detecting edge candidate points in a luminance image.

FIG. 3 shows an example of a brightness value array of a brightness image.

FIG. 4 shows an example of a brightness value array of a brightness image.

FIG. 5 shows an example of an edge candidate point image.

FIG. 6A is an explanatory diagram of creating a histogram of points at which a straight line connecting two points in an edge candidate point image and an end point of the image intersect with each other in the first embodiment. (B) shows an intersection histogram created in Example 1.

FIG. 7 shows an example of an edge candidate point image.

FIG. 8A is an explanatory diagram of a line segment representing a set of points in an edge candidate point image according to the second embodiment. (B) shows an intersection histogram created in Example 2.

FIG. 9A is an explanatory diagram of a line segment representing a set of points in an edge candidate point image according to the third embodiment. (B) shows an intersection histogram α created in the third embodiment.

FIG. 10A is a diagram for explaining creation of a line segment and an intersection histogram β representing a set of points in an edge candidate point image according to the third embodiment. (B) shows an intersection histogram β created in Example 3.

FIG. 11A is a diagram for explaining generation of a line segment and an intersection histogram β representing a set of points in an edge candidate point image in the fourth embodiment. (B) shows an intersection histogram β created in the fourth embodiment.

FIG. 12A shows an intersection histogram α created at the time of extracting the first line segment in the fifth embodiment. (B) In Example 5, an intersection histogram γ obtained by modifying the intersection histogram α is shown.

FIG. 13A shows an intersection histogram α created at the time of extracting the first line segment in the sixth embodiment. (B) In the sixth embodiment, the intersection histogram β created when the first line segment is extracted is shown. (C) An intersection histogram δ obtained by subtracting the intersection histogram β from the intersection histogram α in Example 6 is shown.

FIG. 14A is a diagram for explaining generation of a line segment representing a set of points in an edge candidate point image and an intersection histogram β according to the seventh embodiment. (B) shows an intersection histogram β created in Example 7.

FIG. 15A shows an intersection histogram α at the stage when the second line segment is extracted in the seventh embodiment. (B) shows the remaining histogram obtained by deleting the histogram corresponding to the second line segment from the intersection histogram α.

FIG. 16A is a diagram for explaining creation of a histogram of coordinates of points on a line segment or within a certain distance of the line segment in the eighth embodiment. (B) A histogram of coordinates of points on a line segment created in Example 8 is shown.

[Explanation of symbols]

 100 Color Image Input Device (Color Scanner) 102 Luminance Calculation Unit 104 Edge Candidate Point Extraction Unit 106 Line Segment Extraction Unit

Claims (10)

[Claims] 1. A method for extracting from a set of points in an image, one or more line segments representing the set, the method comprising:
A feature is that a histogram of points that intersect the end points of the image is created for all straight lines connecting points, and a straight line connecting the point having the maximum value of the histogram and the point having the second largest value is extracted as a line segment. The line segment extraction method. 2. A method for extracting one or more line segments representing a set from a set of points in a polygonal image surrounded by a plurality of sides, and connecting all two points in the image. Create a histogram of the points that intersect the end points of the image for the straight line, and use the straight line that connects the point that takes the maximum value of the created histogram and the point that takes the maximum value of the histogram for the side that does not include the line A line segment extraction method characterized by extracting. 3. A method for extracting one or more line segments representing a set from a set of points in the image, the method comprising:
Create a histogram α of points that intersect the end points of the image for all straight lines connecting points, and create a histogram β of points that intersect the end points of the image for all straight lines that pass through the points that take the maximum value of this histogram α. A line segment extracting method, characterized in that a straight line connecting a point having the maximum value and a point having the second largest value of the histogram β is extracted as a line segment. 4. A method for extracting one or more line segments representing a set from a set of points in a polygonal image surrounded by a plurality of sides, and connecting all two points in the image. Create a histogram α of points that intersect the end points of the image for the line
Create a histogram β of points that intersect the end points of the image of all the straight lines that pass through the point where the maximum value of this histogram α is obtained, for the side that does not include this point, and set the maximum value of the histogram α and the maximum of the histogram β. A line segment extracting method, which is characterized in that a straight line connecting a value-taking point is extracted as a line segment. 5. The line segment extraction method according to claim 3, wherein the histogram corresponding to the extracted line segment is deleted from the histogram α created when one line segment is extracted, and the remaining histogram A line segment extraction method characterized in that the next line segment is extracted by using the same method as the previously extracted line segment. 6. The line segment extraction method according to claim 3 or 4, wherein the histogram β is subtracted from the histogram α created when one line segment is extracted, and the remaining histogram is used as the histogram α, and the next line A line segment extraction method, characterized in that a segment is extracted by the same method as the line segment extracted previously. 7. The line segment extracting method according to claim 5, wherein when the histogram corresponding to the line segment extracted from the histogram α is deleted and the maximum value of the remaining histogram is smaller than a predetermined threshold value, the line is extracted. A line segment extraction method characterized by terminating the extraction process of the minutes. 8. The line segment extracting method according to claim 6, wherein when the maximum value of the remaining histogram obtained by subtracting the histogram β from the histogram α becomes smaller than a predetermined threshold value, the line segment extracting process is terminated. A line segment extraction method characterized by the above. 9. The line segment extracting method according to claim 4, wherein for a side including a point having the maximum value in the histogram α, the histogram corresponding to the extracted line segment is deleted from the histogram α, and the remaining histogram Is again used as the histogram α, and for the side that does not include the point having the maximum value of the histogram α, the histogram β is subtracted from the histogram α, and the remaining histogram is used again as the histogram α, and the next line segment is extracted before. A line segment extraction method characterized by extracting in the same manner as a line segment. 10. The line segment extracting method according to claim 1, wherein a histogram of coordinates of points in the image corresponding to the extracted line segment is created, and the end points of the line segment are extracted from the histogram. A line segment extraction method characterized by detecting.