JP2001052167A - Image processing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image processing method capable of reducing an arithmetic quantity and performing high speed processing in spite of highly accurate processing in the case of template matching based on mutual correlation. SOLUTION: Concerning the image processing method using template matching based on mutual correlation, this method has a step for segmenting and fixing an input image from a fetched image through loose position detection and a step for detecting a maximum correlation position by comparing the input image with a template image while successively changing a segmenting position on the side of the template image. At the time of detecting the maximum correlation position, processing is performed for equalizing a luminance average value and a luminance dispersion for template images to be successively segmented and the peak of a mutual correlation value is detected by calculating the total sum of products of respective pixels in the input image and the template image.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing method capable of performing image alignment, extraction of a specific pattern, and the like using pattern matching by cross-correlation in an automated manufacturing line or the like.

[0002]

2. Description of the Related Art Manufacturing lines for assembly, mounting, inspection, etc.
When aligning workpieces flowing through
Pattern matching method is used. Cross-correlation
In the case of switching, the image captured by the camera as shown in FIG.
1 and the template 1 as a reference image
Input image corresponding to the size (N1 × N1) of the port image T
I ₁ ~ I_Three Are sequentially cut out and compared with the template image T.
Is performed. That is, the mutual phase as shown in equation (1)
Mutual phase of input image I and template image T
The function C (a, b) is sequentially obtained. The coordinates (a, b)
Represents the upper left position of the input image in the m and n directions.

[0003]

(Equation 1)

As the similarity between the input image I and the template image T increases, the cross-correlation value C (a, b) increases,
Approaches one. The maximum correlation position is the upper left position (a, b) of the clipped input image at which the expression (1) is maximized. Where I is the average luminance of the input image, T is the average luminance of the template image, Iσ is the sum of the squares of the difference between the luminance of each pixel of the input image and the average luminance, and Tσ is the average of the luminance of each pixel of the template image. This is the sum of the square of the difference from the luminance.

FIG. 2 is a graph showing the cross-correlation value C. In the case of FIG. 2, the input image I ₂ has the largest correlation value C
Is high, indicating that this is the maximum correlation position.

[0006]

However, when calculating equation (1), I _{(a, b)} (m,
n)-The calculation of I and I [ sigma] is required every time the cutout position (extraction position) of the input image I is changed. That is,
Every time the extraction position of the input image I is changed, the cross-correlation value C
Since the calculation of (a, b) needs to be performed sequentially, it takes a lot of time to detect a peak point. Since the processing time after capturing the image affects the processing capability of the positioning equipment and the inspection equipment using the image processing, it is necessary to shorten the processing time as much as possible.

SUMMARY OF THE INVENTION An object of the present invention is to provide an image processing method capable of reducing the amount of calculation and performing high-speed processing despite high-accuracy processing in template matching by cross-correlation. .

[0008]

According to an aspect of the present invention, there is provided an image processing method using template matching based on cross-correlation, comprising the steps of: fixing an input image; Detecting the maximum correlation position by comparing the input image with the template image while sequentially changing the cut-out position, thereby providing an image processing method. In the case of the template matching based on the conventional cross-correlation, there are many calculations required for the cross-correlation only after the image is captured. Therefore, it takes a lot of time to calculate the cross-correlation value C, which causes a reduction in processing capacity. On the other hand, according to the first aspect of the present invention, since the input image is fixed and the cutout position of the template image is changed, most of the cross-correlation values required for calculation can be calculated in advance by external setup. The number of calculations after image capture can be reduced. As a result, the processing capacity can be significantly improved.

According to a second aspect of the present invention, in the image processing method using template matching based on cross-correlation, one of the input image and the template image is fixed, and the other of the input image and the template image is cut out. Detecting the maximum correlation position by comparing the input image and the template image while sequentially changing the luminance value of the input image or the template image that is sequentially cut out when the maximum correlation position is detected. In addition, the present invention provides an image processing method for performing processing for making the same luminance variation and detecting a peak of a cross-correlation value by calculating a sum of products of pixels of an input image and a template image. The expression for obtaining the cross-correlation requires a great deal of calculation, but if only the peak of the cross-correlation value is to be found, a simplified expression based on the sum of the products of the pixels of the input image and the template image may be used. However, in the simplified calculation, if there is a change in the average luminance value or the luminance variation of each pixel of the sequentially cut out image, a large difference from the cross-correlation formula appears. Therefore, by making the average luminance value and the luminance variation the same, it is possible to find the peak of the cross-correlation value at high speed only by calculating the sum of the products of the pixels of the input image and the template image. Things.

Further, the invention according to claim 3 is a combination of claims 1 and 2, wherein in the image processing method using template matching by cross-correlation, a step of fixing an input image; Detecting the maximum correlation position by comparing the input image and the template image while sequentially changing the extraction position on the template image side. When detecting the maximum correlation position, the luminance of the template image to be sequentially extracted is determined. This is an image processing method that performs processing for making the average value and the luminance variation the same, and detects the peak of the cross-correlation value by calculating the sum of the products of the pixels of the input image and the template image. In this case, the effects of the first and second aspects can be obtained.

According to a fourth aspect of the present invention, in the image processing method according to the second or third aspect, the processing for equalizing the luminance average value and the luminance variation includes the steps of: It is characterized by processing to a value. That is, when the image has a feature portion having a different luminance from other portions in the image, the portion other than the feature portion has a constant brightness value, so that only the sum of the products of the respective pixels of the input image and the template image is calculated. The maximum correlation position can be easily detected.

According to a fifth aspect of the present invention, in the image processing method according to the second or third aspect, the processing for equalizing the luminance average value and the luminance variation is performed when a current image is sequentially cut out. It is characterized in that part of the end of the image is overwritten on another part of the next cut-out image. In this case as well, by overwriting a part other than the characteristic part with another part of the next cut-out image, the data amount can be reduced and the maximum correlation position can be easily detected.

[0013]

3 and 4 show a first embodiment of the flow of image processing according to the present invention. Hereinafter, the flow of the image processing of FIG. 3 will be described with reference to FIG. First, the template image T is created and registered slightly larger than the cut-out range of the input image I, which will be described later, including the search area (step S1). Here, the template image T is an image including the mark portion 10 and the other portion 11. Cutting a plurality of partial template image T ₁ through T ₃ of the same size as the input image I from the template image T (Step S2). Of these, the partial template image T
_The mark 10 is located at the center of ₃ . Next, for each of the partial template images T _{1 to} T ₃ , the cross-correlation value C
In the calculation formula of (a, b), ΔT (m, n) −
T ｝ and the denominator Tσ are calculated in advance, stored in a table, and stored (step S3). These steps S1 to
The operation of S3 is performed in an external setup separately from the capture of the input image.

On the other hand, an image 12 of the object is captured by an imaging device such as a CCD camera (step S4). Next, coarse position detection is performed from the captured image 12 (step S5), and an input image I in which the mark 10 is likely to exist is detected from the captured image 12. Note that the mark 10 does not need to be located at the center of the input image I. As a method of detecting the coarse position, a known method of measuring the center of gravity, matching by the residual method, or the like may be used. Next, only one input image I whose coarse position is detected is cut out and fixed (step S6).
Then, among the equations for calculating the cross-correlation value C (a, b), {I _{(a, b)} (m, n) -I } of the numerator and Iσ of the denominator are calculated from the cut-out input image I ( Step S7). This calculation only needs to be performed once.

Next, ΔT tabulated as described above
(M, n) −T } and Tσ, and the calculated {I _{(a, b)}
From (m, n) -I } and Iσ, a cross-correlation value C (a, b) is calculated by equation (1) (step S8).
Then, the maximum correlation position is calculated from the maximum value of the cross-correlation value C (a, b) (step S9). In the example of FIG. 4, the template image T ₃ is located at the maximum correlation position with the input image I.
It is. Thereafter, it may be obtained center position of the mark 10 by back calculation the position of the template image T ₂ from the template image T _3.

As described above, in the equation for calculating the cross-correlation value C (a, b), {I _{(a, b)} (m, n) -I } of the numerator and Iσ of the denominator are the cut-out input values. The calculation is performed only once for the image I, and {T (m, n) -T } and the denominator Tσ of the numerator can be calculated only in advance in the search range and tabulated. The subsequent processing is simplified, and the cross-correlation value C (a,
b) can be calculated.

FIG. 5 shows a second example of the image processing method according to the present invention.
1 shows an embodiment. In order to obtain the maximum correlation position using the cross-correlation equation such as the equation (1), the average luminance I and the variation Iσ
Must be calculated, which takes a long calculation time. On the other hand, if only the relative similarity peak position is found, the following simplified expression (2) can be considered.

[0018]

(Equation 2) Note that α and B are constants.

In the case of sequentially comparing the input image with the template image while changing the cutout position of the input image as in the related art,
Average luminance I and luminance variation Iσ of cut out input image
Equation (2) holds only when is constant. However, actually, the input image has luminance unevenness and cannot be said to be constant. Therefore, the peak position cannot be detected by the simplified formula (2).

On the other hand, in the method of cutting out and fixing only one input image and comparing the input image and the template image while sequentially changing the cutting position of the template image, the average luminance I of the input image is reduced. And the luminance variation Iσ are constant, and by performing pre-processing to make the average luminance value and the luminance variation of the template image the same, the average luminance T and the luminance variation Tσ of the template image are obtained.
Can also be constant. Thus, the relative peak position can be detected using the simplified expression (2).

Average luminance and luminance of template image
Here, as a preprocessing method to make the variation the same,
Next template image T₁ ~ T_Three Constant around
Fill the luminance value (here the gray part)
You. In this case, the input image and template image
Only the sum of elementary products needs to be calculated, and the ΔT
(M, n)-TIt is not necessary to calculate｝ and Tσ in advance
In addition, there is no need to create a table. Also, regarding the input image
｛I after coarse position detection _{(a, b)} (M, n)-I｝ And Iσ
Is unnecessary. Peak of the simplified formula P (a, b)
If the position is detected, the peak position of the cross-correlation value C (a, b)
And the maximum correlation position can be obtained. Figure
In the case of No. 5, the template at the maximum correlation position with the input image I
Rate image is T_Three Can be easily detected.

FIG. 6 shows a third embodiment of the image processing method according to the present invention.
1 shows an embodiment. In FIG. 5, the peripheral portion of the template image T is painted with a constant luminance in order to make the average luminance value and the luminance variation of the template image the same. This is to overwrite another end 16 of the next cutout image. That is, the end 15 of the currently cut out template image T _n is inverted and overwritten with another portion 16 of the next cut out template image T _{n + 1} . Note that, in the drawing, the end 1
Although the example in which 5 is L-shaped is shown, it may be a rectangular end. In this way, the peak position can be detected using the simplified matching calculation formula P (a, b), as in the case of FIG.

In the method of overwriting the end 15 other than the mark 10 of the current cut image on another end 16 of the next cut image as in the third embodiment, the target image has uneven brightness. Since there is no problem, the image to be overwritten need not be a template image, but may be an input image. Therefore, in the case of the third embodiment, not only the case where the input image is fixed and the cutout position of the template image is changed, the template image may be fixed and the cutout position of the input image may be changed.

[0024]

As is apparent from the above description, claim 1
According to the invention of the above, since the input image is fixed and the cutout position of the template image is changed, most of the cross-correlation value required can be calculated in advance by external setup, and the number of operations after image capture is reduced. can do. as a result,
The calculation of the cross-correlation value can be performed at high speed, and the processing capability can be greatly improved. According to the second aspect of the invention, when the maximum correlation position is detected, the average luminance value and the luminance variation are set to be the same, and a simplified expression based on the sum of the products of the pixels of the input image and the template image is used. The peak position of the correlation value can be obtained with high accuracy and at high speed. Further, since the third aspect of the present invention has the features of the first and second aspects of the present invention, the positioning can be performed at higher speed and with higher accuracy.

[Brief description of the drawings]

FIG. 1 is an explanatory diagram showing a conventional pattern matching method based on cross-correlation.

FIG. 2 is a diagram illustrating a relationship between an input image and a cross-correlation value.

FIG. 3 is a flowchart of a first embodiment of an image processing method according to the present invention.

FIG. 4 is an explanatory diagram showing a method for processing an input image and a template image in the image processing method shown in FIG. 3;

FIG. 5 is an explanatory diagram showing a processing method according to a second embodiment of the image processing method according to the present invention.

FIG. 6 is an explanatory diagram showing a processing method according to a third embodiment of the image processing method according to the present invention.

[Explanation of symbols]

 I Input image T Template image 10 Mark

 ────────────────────────────────────────────────── ─── Continued on the front page (72) Inventor Yoshinori Ikeda 2-26-10 Tenjin, Nagaokakyo-shi, Kyoto F-term in Murata Manufacturing Co., Ltd. 5L096 AA06 BA05 DA02 EA35 FA32 FA34 FA69 GA41 HA08 JA09 JA11

Claims (5)

[Claims] In an image processing method using template matching based on cross-correlation, a step of fixing an input image and a step of comparing the input image and the template image while sequentially changing a cutout position on the template image side, Detecting the maximum correlation position. 2. An image processing method using template matching based on a cross-correlation, wherein one of an input image and a template image is fixed, and the other of the input image and the template image is sequentially changed while cutting out the input image and the template image. Comparing the template image with the template image to detect the maximum correlation position. When the maximum correlation position is detected, a process for making the average luminance value and the luminance variation of the input image or the template image cut out sequentially the same is performed. An image processing method for detecting a peak of a cross-correlation value by calculating a sum of products of pixels of an input image and a template image. 3. An image processing method using template matching based on cross-correlation, wherein the step of fixing the input image and the step of comparing the input image and the template image while sequentially changing the cutout position on the template image side, Detecting a maximum correlation position,
When the maximum correlation position is detected, a process of making the average luminance value and the luminance variation of the template image sequentially cut out the same is performed, and the peak of the cross-correlation value is detected by calculating the sum of the products of the respective pixels of the input image and the template image. An image processing method comprising: 4. The method according to claim 2, wherein the processing for equalizing the luminance average value and the luminance variation is to process a portion other than a common portion of the sequentially cut out image to a constant luminance value. Image processing method. 5. The processing for equalizing the brightness average value and the brightness variation is to overwrite a part of the end of the current cut image with another part of the next cut image when cutting out the image sequentially. The image processing method according to claim 2 or 3, wherein