JPH0778832B2 - Edge detection method - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention utilizes an image density difference between an object and a background, and an edge which is a boundary line between the object and the background based on a binary image binarized using a predetermined threshold value. The present invention relates to an edge detection method for detecting the.

[0002]

2. Description of the Related Art Conventionally, as an edge detecting method for detecting an edge which is a boundary line between an object and a background, a differential value for each pixel of a grayscale image is obtained and binarized based on the obtained differential value. A method for correcting the threshold value for performing the above is disclosed in JP-A-64-7.
It is disclosed in Japanese Patent Publication No. 9882.

This method is based on the knowledge that the maximum differential value is small when the density difference between the object and the background is small, and the maximum differential value is large when the density difference is large. That is, by setting the threshold for binarization higher as the maximum value of the differential value increases,
An appropriate binary image is obtained so that the object and the background can be easily distinguished.

[0004]

The above method has the advantage that an appropriate threshold value can be set according to the magnitude of the contrast between the object and the background. However, at least for the area including the edge, it is necessary to calculate the differential value for all pixels, and therefore it is necessary to perform a large amount of calculation on a grayscale image in which each pixel is multivalued, and the calculation amount is very large. There was a problem of becoming.

The present invention is intended to solve the above-mentioned problems, and requires a relatively small amount of calculation for a binary image.
It is intended to provide an edge detection method capable of setting an appropriate threshold value for binarization.

[0006]

In [SUMMARY OF] of claim 1 the invention, in order to achieve the above object, the edge detection method for detecting an edge which is a boundary line between the object and the background based on the binary image, detecting Aligned in the direction orthogonal to the edge to be
White and black pixels on multiple lines along the edge
Find the number of concatenated one pixel of
Therefore, the first line where the number of concatenated pieces is less than the specified number
There is a second line adjacent to
And on the opposite side of the first line to the second line
When there is a third line whose number is equal to or larger than the number,
In is detected as a straight line that is a candidate including an edge, a pair of windows is set across this straight line, and whether or not the above straight line includes an edge is determined based on the number of white pixels or black pixels included in each window. If it is determined that the edge is not included, the threshold for binarization is changed to 2
A value image is created again, and the above-described processing is applied to the newly created binary image. In the invention of claim 2,
In order to achieve the above objective,
Edge detection that detects edges that are boundaries between elephants and the background
In the output method, the candidate straight line including the edge is detected.
Then, one line of the same shape containing the same number of pixels across this straight line
Set a pair of windows, and the white image included in each window
Amount of light emitted to the object based on the number of elementary or black pixels
And the irradiation light by judging the contrast between the object and the background
If the amount and contrast are sufficient, the straight line
Edge is not included and it is determined that edge is not included.
Change the threshold for binarization and recreate the binary image.
Then, the above processing is applied to the newly created binary image.
Of.

[0007]

According to the method of the invention of claim 1 , the error to be detected
Multiple pieces along the edge arranged in the direction orthogonal to the edge
One of the white and black pixels on the line
Determine the number of concatenated pixels and specify the number of concatenated pixels
Adjacent to the first line that is less than the number
There is a second line that becomes and for the second line
On the opposite side of the first line, the
A candidate that contains an edge on the second line when the in exists
Differentiation is performed by detecting as a straight line
Find the number of connected white or black pixels without
You can obtain straight lines that are candidates for
It is. That is, every pixel in a multi-value grayscale image
Compared to the case of performing the differential operation on
Significantly less calculation is required to obtain candidate straight lines
It will disappear. Moreover, after detecting a straight line that is a candidate including an edge in a binary image, it is determined whether or not the straight line includes an edge by detecting the distribution of white pixels or black pixels on both sides of this straight line. Therefore, if the threshold value is properly set and the straight line includes an edge, the straight line includes an edge due to a large difference in the number of white pixels or black pixels across the straight line. Whether or not it can be accurately determined.
Further, when it is determined that the edge is not included, the threshold value is changed, the binary image is created again, and then the above process is repeated. Therefore, the threshold value is changed until the threshold value becomes an appropriate value. Therefore, the appropriate threshold value can be converged to an appropriate threshold value, and the appropriate threshold value can be obtained with a relatively small amount of calculation for only the binary image. Contract
According to the method of the invention of claim 2, straight lines that are candidates for edges
A pair of windows of the same shape that include the same number of pixels across
And set the white pixels or black pixels included in each window.
Based on the number of pixels of
The amount of irradiation light and
When the last is sufficient, the above straight line is determined to include an edge.
By inserting the straight line that is a candidate for the edge,
The image of interest among the same number of pixels in the shape window
Depending on the number of pixels (white pixels or black pixels), the target object is illuminated.
Determine the amount of light and the contrast between the object and the background
The edge can be calculated with a small amount of calculation without relying on the differential calculation.
It is possible to determine a straight line including. Moreover,
Threshold for binarization when it is determined that no edge is included
Is changed to create a binary image again, and a new image is created.
Since the above-mentioned processing is applied to the binary image, the invention of claim 1
In the same manner as above, proper binarization can be performed.

[0008]

EXAMPLE As shown in FIG. 2, an image including an object is C
The image is captured by the image input device 1 such as a CD camera. An image signal which is an analog signal output from the image input device 1 is converted into a digital signal through an A / D conversion circuit 2 and then binarized by a binarization processing unit 3 and stored in a binary image memory 4. To be done. The threshold value in the binarization processing unit 3 is set by the threshold value setting unit 5.

In the present invention, an appropriate threshold value is set in the binarization processing unit 3 based on the binary image stored in the binary image memory 4. As shown in FIG. A straight line (hereinafter, abbreviated as an edge candidate) that is a candidate including an edge that is a boundary line between the background and the background is detected (steps 100 and 101). That is, in the edge detection unit 6,
Regarding the binary image stored in the binary image memory 4, FIG.
As shown in, the mask M is set in advance as an area where an edge should be detected, and an edge candidate is detected for the pixel P in the mask M. For example, in order to detect the edge candidate in the vertical direction, a sequential search is performed in the horizontal direction. First, the pixel value of the leftmost vertical line L ₁ in the mask M is sequentially determined from the upper end to the lower end. To do. Here, the value of the white pixel (open circle) is 0, the value of the black pixel (circle with diagonal lines) is 1, and there are two or more lines L _{1 to} L ₅ connecting three or more black pixels. When connected (here, the lines L ₃ and L ₄ satisfy the condition), the leftmost line L ₃ of the connected lines L ₃ and L ₄ is set as an edge candidate. To find horizontal edge candidates,
A sequential search may be performed in the vertical direction. Further, here, it is considered that the inside of the object becomes a black pixel, but when the inside of the object becomes a white pixel, the edge candidate may be detected by connecting the white pixels. The size of the mask M, the number of connected pixels P, the number of connected lines L _{1 to} L ₅ , and the like are not limited to those in the embodiment.

After detecting the edge candidate E as described above, the counter 7 sets a pair of windows W ₁ and W ₂ with the edge candidate E sandwiched in the mask M, as shown in FIG. . Both windows W _1, W ₂ is set to include the same number of pixels P a same shape, each window W _1, W
The number of white pixels or black pixels in the ₂ are counted respectively (step 102 in FIG. 1). For example, when the number of black pixels in each of the windows W ₁ and W ₂ is counted when the inside of the object is a black pixel, the number of black pixels on the object side with respect to the edge candidate E is the number of black pixels on the background side. Since the number of black pixels is significantly larger than the number of black pixels, the difference or ratio of the number of black pixels in both windows W ₁ and W ₂ can be determined to determine whether the edge candidate E is a true edge. It will be.

That is, the judging section 8 compares the difference or ratio of the number of black pixels or white pixels in each of the windows W ₁ and W ₂ counted by the counting section 7 with a predetermined reference value, and compares it with the reference value. It is determined whether or not the edge candidate E is a true edge based on the magnitude relationship. Here, the numbers of black pixels in the windows W ₁ and W ₂ are w1 and w2, and two types of reference values α and β are set. As shown in steps 103 and 104 of FIG. 1, if w2 ≧ α, it is determined that the light amount is sufficient, and if w2-w1 ≧ β, the detected edge candidate E is a true edge because the contrast is large. Is determined (step 105). On the other hand, w2
If <α, it is determined that the light amount is insufficient, and the threshold value setting unit 5 is controlled so as to lower the threshold value for binarization (step 106). Is changed, imaging is performed again, and the above processing is repeated. Even if w2 ≧ α holds, if w2-w1 <β,
The threshold value setting unit 5 is controlled so as to increase the threshold value when it is determined that the light amount is too large (step 107), the imaging is performed again, and the above processing is repeated. Here, the threshold value of the binarization processing unit 3 is set in multiple steps within a predetermined width, and the threshold value is raised or lowered step by step.

In the above embodiment, when the threshold value is changed, the image input device 1 performs image pickup again. However, the picked-up image is stored in the grayscale image memory and the grayscale value is changed when the threshold value is changed. The grayscale image stored in the image memory may be binarized. Further, the binarization processing unit 3, the threshold setting unit 5, the edge detection unit 6, the counting unit 7, and the determination unit 8 can be easily realized by using a microcomputer.

[0013]

According to the invention of claim 1, an edge to be detected is detected.
Multiple lines in the direction along the edges arranged in the orthogonal direction
One of the white and black pixels
Then find the number of concatenated pieces, and the number of concatenated pieces is
Adjacent to the first line, which is less
There is a second line and the first to the second line
There is a third line on the opposite side of
Second line becomes a candidate including edge when present
Since it is detected as a straight line, white
Simply find the number of contiguous pixels or black pixels
A straight line that is a candidate for
Differentiate all pixels in the grayscale image
Compared with the case, the process of finding the straight line that is a candidate for the edge
It has the advantage of significantly reducing the amount of calculation required
It Moreover, after detecting a straight line that is a candidate including an edge in a binary image, it is determined whether or not the straight line includes an edge by detecting the distribution of white pixels or black pixels on both sides of this straight line. Therefore, if the threshold value is properly set and the straight line includes an edge, the straight line includes an edge due to a large difference in the number of white pixels or black pixels across the straight line. Whether or not it can be accurately determined. Further, when it is determined that the edge is not included, the threshold value is changed, the binary image is created again, and then the above process is repeated. Therefore, the threshold value is changed until the threshold value becomes an appropriate value. Can be made to converge to an appropriate threshold value.
This has an advantage that an appropriate threshold value can be obtained with a relatively small amount of calculation of only the value image. That is, the amount of light around the edge can be determined, and the threshold value can be corrected with respect to the change in contrast due to the change in the environment. interest that can be detected
There is a point. The invention of claim 2 is a straight line which is a candidate for an edge.
A pair of windows of the same shape that include the same number of pixels across
And set the white pixels or black pixels included in each window.
Based on the number of pixels of
The amount of irradiation light and
Determine if the straight line including the edge when the last is sufficient
Since the same shape is inserted between straight lines that are candidates for edges,
Number of pixels of interest among the same number of pixels in the window
Depending on the amount of light illuminating the object and the
It is possible to determine the trust, without relying on the differential operation.
It is possible to determine straight lines including edges with a small amount of calculation.
Has the advantage of Moreover, it is determined that the edge is not included.
If it is set, the binary image is changed by changing the threshold value for binarization.
Again for the newly created binary image
Since the processing is performed, an appropriate binary value is obtained as in the first aspect of the invention.
It can be transformed.

[Brief description of drawings]

FIG. 1 is an explanatory diagram of an operation of an embodiment.

FIG. 2 is a block diagram of an embodiment.

FIG. 3 is an operation explanatory diagram relating to the edge detection unit of the embodiment.

FIG. 4 is an operation explanatory diagram of a counting unit according to the embodiment.

[Explanation of symbols]

 1 Image Input Device 2 A / D Conversion Circuit 3 Binarization Processing Unit 4 Binary Image Memory 5 Threshold Setting Unit 6 Edge Detection Unit 7 Counting Unit 8 Judgment Unit

Claims (2)

[Claims] 1. An edge detection method for detecting an edge which is a boundary line between an object and a background based on a binary image,
Along the edges arranged in the direction orthogonal to the edge to be detected
Of white pixels and black pixels on multiple lines in the same direction
Find the number of concatenated one of the viewed pixels and connect
Adjacent to the first line where the number of
There is a second line that exceeds the specified number, and the second line
More than the specified number on the opposite side of the first line to the line
When the second line is present, the second line is detected as a straight line that is a candidate including an edge, and a pair of windows is set with the straight line sandwiched between the white line and the white pixel included in each window. Alternatively, it is determined whether or not the straight line includes an edge based on the number of black pixels, and if it is determined that the straight line does not include an edge, the threshold for binarization is changed to recreate a binary image, and a new image is created. An edge detecting method, characterized in that the above-mentioned processing is performed on the binary image created in (1). 2. A boundary between an object and a background based on a binary image.
In the edge detection method for detecting edges that are field lines,
After detecting a straight line that is a candidate including edges,
A pair of windows of the same shape containing the same number of pixels sandwiched between
Set the white pixel or black pixel included in each window.
Based on the number of pixels, the amount of light emitted to the object and the object and spine
The contrast with the scene is determined to determine the irradiation light amount and contrast.
It is determined that the above straight line includes an edge when the strike is sufficient.
However, if it is determined that the edge is not included, binarization is performed.
Create a new binary image by changing the threshold and creating a new one.
It is characterized in that the above-mentioned processing is applied to the extracted binary image.
Edge detection method.