JP4467690B2 - Template setting method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention belongs to the technical field of setting a template in template matching. In particular, the present invention relates to a template setting method for setting a template for detecting a positional deviation amount between a reference image and an inspection target image in a printed matter inspection apparatus.
[0002]
[Prior art]
2. Description of the Related Art There has been known a printed matter inspection apparatus that compares a reference image that is a captured image of a printed matter that is an inspection reference and an inspection target image that is a captured image that is an inspection target to inspect the quality of the printed matter. In this type of inspection apparatus, when there is a positional deviation between the reference image and the inspection target image, if the two are compared as they are, a malfunction may occur in which a non-defective printed matter is determined as a defective product. In order to avoid this malfunction, the inspection level (inspection level) must be lowered accordingly. Therefore, in order to improve the inspection performance in this inspection apparatus, the positional deviation between the reference image and the inspection target image is corrected and compared.
[0003]
In order to correct this misregistration, it is necessary to acquire the misregistration amount. For this purpose, template matching (pattern matching) is calculated. For example, the degree of coincidence is calculated by changing the relative positions of the reference image and the inspection target image, and the positional deviation amount is acquired based on the relative position having the highest degree of coincidence.
[0004]
[Problems to be solved by the invention]
However, if the degree of coincidence is calculated by considering almost the entire area of the image as a template, the amount of calculation becomes enormous. In addition, since inspection is normally performed in-line on a printing machine or the like that operates at high speed, it is difficult to realize an inspection device due to time constraints. Therefore, if it can be assumed that the inspection target article has no distortion such as expansion and contraction like a printed matter, the amount of positional deviation is constant regardless of the location. Therefore, even if a part of the image is considered as a template and calculated, it is possible to obtain a result that matches the total amount of displacement.
[0005]
Thus, if a part of the image is calculated as a template, the amount of calculation is reduced, and the degree of coincidence can be calculated at high speed. However, in order to obtain high accuracy and stability equivalent to the case where almost the entire area of the image is used as a template, the selection of the template setting location is crucial. If the template contains only a solid part of the pattern or an edge component in one direction of the pattern, the correction is definitely made.
[0006]
The present invention has been made to solve the above problems. The purpose is to set an appropriate template capable of obtaining high accuracy and stability as a template for calculating the positional deviation amount.
[0007]
[Means for Solving the Problems]
The above problems are solved by the present invention described below. In other words, the template setting method according to the first aspect of the present invention provides an edge extraction process for extracting an edge in a reference image obtained by imaging a printed matter serving as an inspection reference, and predetermined edges for at least two edges in a predetermined direction. Count the number of edge pixels in the template size area at the position, create a position sequence in which the position of the template size area is rearranged in the order of the number of edge pixels in each edge direction, in each edge direction position sequence Comparing the position of the template size region in the order of the number of edge pixels in either direction, and selecting the region of the reference image at the position of the common template size region that appears first as a template. It is what you have. According to the present invention, edges are extracted from a reference image obtained by imaging a printed matter that is an inspection reference in the edge extraction process, and the number of edge pixels in at least two predetermined directions in the template size area is large in the area selection process. The positions of the templates are compared in order, and the region of the reference image that appears first at the common position is selected as the template. That is, since the selected template is an area including many edges in two predetermined directions, the degree of coincidence changes greatly with respect to a two-dimensional change in relative position. Therefore, an appropriate template capable of obtaining high accuracy and stability is set as a template for calculating the positional deviation amount.
[0008]
The template setting method according to claim 2 of the present invention is the template setting method according to claim 1, wherein the region selection step is a step of selecting, as a template, a region where edges in the predetermined direction are connected. It is what I did. According to the present invention, the selected template is an area where edges in a predetermined direction are connected. That is, since the edges in the predetermined direction are closely present, the area size can be small. Therefore, since the amount of calculation is reduced in the calculation of the degree of coincidence, the inspection speed can be increased.
[0009]
The template setting method according to a third aspect of the present invention is the template setting method according to the first or second aspect, wherein the region selection step removes an edge other than the predetermined direction, and then removes the edge in the predetermined direction. This is a process of selecting a region of the reference image including many as a template. According to the present invention, edges other than the predetermined direction are removed from the selected template. Therefore, an appropriate template that can obtain particularly high precision and stability is set.
[0010]
A template setting method according to a fourth aspect of the present invention is the template setting method according to any one of the first to third aspects, wherein the region selection step uses a region having no similar pattern in the misalignment correction region as a template. It is a process of selecting. The misregistration correction area is a range in which the degree of coincidence is calculated by changing the relative position between the template and the image. In other words, if a similar pattern exists in the misregistration correction area, the degree of coincidence is maximized in the vicinity of the similar pattern, and there is a possibility that the misregistration is calculated as a real pattern. Therefore, according to the present invention, it is possible to avoid an error in calculating a misregistration amount by mistaking a similar design and a real design.
[0011]
DETAILED DESCRIPTION OF THE INVENTION
Next, embodiments of the present invention will be described. The process of data processing in the template setting method of the present invention is shown in FIG. In FIG. 1, a reference image is an image obtained by imaging an article (for example, a printed matter) that is a reference for inspection. This reference image is stored in advance in a storage device of a data processing system to which the template setting method of the present invention is applied.
[0012]
First, in step S1 of FIG. 1, a horizontal differentiation process is performed on the reference image. This horizontal differentiation process is a process of calculating a differential coefficient by considering a pixel value as a function value of a pixel position when tracing a pixel in the horizontal direction (row direction) in the reference image (when associating). Since it is actually a digital image, the difference is calculated. Here, since there is no significance in distinguishing the sign of the difference, the absolute value of the difference is calculated. When this horizontal differentiation process is performed on the entire reference image, a horizontal difference image is obtained.
[0013]
In the horizontal difference image obtained by the horizontal differentiation process, the absolute value of the difference is large in the portion where the change in the pixel value in the horizontal direction is large in the reference image, and the absolute value of the difference is small in the portion where the change in the pixel value is small. That is, the horizontal difference image is an image in which the edge in the vertical direction (column direction) in the reference image is emphasized. As schematically shown in (1) in FIG. 1, even if the edge is not a complete vertical edge, if the component is included in an oblique edge, the absolute value of the difference is obtained as it is. Is emphasized.
[0014]
Next, in step S2, the horizontal difference image is binarized to obtain a horizontal binarized difference image. The present invention is not limited by the method of determining the threshold value (boundary value) when performing binarization. For example, a discriminant analysis method in which discriminant analysis in multivariate analysis is applied to threshold determination, a mode method in which pixel values corresponding to valleys in a pixel value histogram of an image are used as threshold values, and the like can be applied. This horizontal binarized difference image can divide the image into two parts based on pixel values. That is, there are two parts: a picture part (for example, pixel value “1”) and a background part (for example, pixel value “0”).
[0015]
Next, in step S3, vertical reduction processing is performed on the horizontal binarized differential image to obtain a vertical edge image. In this vertical reduction process, when pixels are traced in the vertical direction (column direction) in the horizontal binarized difference image (when they are associated), the number of pixels in the pattern portion (for example, the pixel value is “1”) is a predetermined number of pixels. Is not continuous, it is a process of replacing it with a background portion (pixel value is set to “0”). That is, when the number of pixels having a pixel value “1” continuous in the vertical direction is less than the predetermined number of pixels, the pixel is erased and has a pixel value “0”. As the predetermined number of pixels, in the vertical reduction process, the number of pixels having a size that eliminates the oblique edges and leaves the vertical edges is applied.
[0016]
Next, in step S4, the vertical edge image is scanned with a predetermined template size area at a predetermined position (or at a predetermined interval), and the number of vertical edge pixels at that position (address) is counted. Save the number of vertical edge pixels.
Next, in step S5, the aforementioned positions are rearranged (sorted) in the order of the number of vertical edge pixels, and the position string is stored as a vertical position string.
[0017]
The processing process starting from the horizontal differentiation process (S1) for the reference image has been described above with reference to FIG. Next, although it is the same, the process starting from the vertical differentiation process (S6) for the reference image will be described.
First, in step S6 in FIG. 1, vertical differentiation processing is performed on the reference image. This vertical differentiation process is a process of calculating a differential coefficient by considering a pixel value as a function value of a pixel position when tracing (associating) a pixel in the vertical direction (column direction) in the reference image. Since it is actually a digital image, the difference is calculated. Here, since there is no significance in distinguishing the sign of the difference, the absolute value of the difference is calculated. When this vertical differentiation process is performed on the entire reference image, a vertical difference image is obtained.
[0018]
In the vertical difference image obtained by this vertical differentiation process, the absolute value of the difference is large in the portion where the change in the pixel value in the vertical direction is large in the reference image, and the absolute value of the difference is small in the portion where the change in the pixel value is small. That is, the vertical difference image is an image in which the edge in the horizontal direction (row direction) in the reference image is emphasized. As schematically shown in (6) of FIG. 1, even if the edge is not a complete horizontal edge, if the component is included in an oblique edge, the absolute value of the difference is obtained as it is, so that the edge is obtained. Is emphasized.
[0019]
Next, in step S7, the vertical difference image is binarized to obtain a vertical binarized difference image. The present invention is not limited by the method of determining the threshold value (boundary value) when performing binarization. For example, a discriminant analysis method in which discriminant analysis in multivariate analysis is applied to threshold determination, a mode method in which pixel values corresponding to valleys in a pixel value histogram of an image are used as threshold values, and the like can be applied. This vertical binarized difference image can divide the image into two parts based on pixel values. That is, there are two parts: a picture part (for example, pixel value “1”) and a background part (for example, pixel value “0”).
[0020]
Next, in step S8, horizontal reduction processing is performed on the vertical binary difference image to obtain a horizontal edge image. In this horizontal reduction process, when pixels are traced in the horizontal direction (row direction) in the vertical binarized difference image (when they are associated), the picture portion (for example, the pixel value is set to “1”) has a predetermined number of pixels. Is not continuous, it is a process of replacing it with a background portion (pixel value is set to “0”). That is, when the number of pixels having a pixel value “1” continuous in the vertical direction is less than the predetermined number of pixels, the pixel is erased and has a pixel value “0”. As the predetermined number of pixels, in the horizontal reduction process, the number of pixels having a size that eliminates the oblique edges and leaves the horizontal edges is applied.
[0021]
Next, in step S9, a region having a predetermined template size is scanned at a predetermined position (or at a predetermined interval) on the horizontal edge image, and the number of horizontal edge pixels at the position (address) is counted. Save the number of horizontal edge pixels.
Next, in step S10, the above-mentioned positions are rearranged (sorted) in the order of the number of horizontal edge pixels, and the position string is stored as a horizontal position string.
[0022]
The processing process starting from the vertical differentiation process (S6) for the reference image has been described above with reference to FIG. Next, the process (step S11) for determining the position (address) of the template to be actually used will be described.
In step S11, in the vertical position sequence in the order of the number of vertical edge pixels and the horizontal position sequence in the order of the number of horizontal edge pixels, the position of the template in the order of which the number of vertical edge pixels or the number of horizontal edge pixels is larger. Compare Then, the common position that appears first is selected as the template position.
[0023]
More specifically, the positions of the first column (first column) in the vertical position column and the horizontal position column are compared. If it is the same position, that position is selected as the template position. If not the same position, the positions of the first column (first column) and the second column in the vertical position column and the horizontal position column are compared. If a common position is included in the first column (first column) and the second column, that position is selected as the template position. If the common position is not included, the positions of the first column (first column) to the third column in the vertical position column and the horizontal position column are compared. If a common position is included in the first column (first column) to the third column, that position is selected as the template position. If the common position is not included, the positions of the first column (first column) to the fourth column in the vertical position column and the horizontal position column are compared. If a common position is included in the first column (first column) to the fourth column, that position is selected as the template position. Thus, the comparison is continued until a common position is included.
[0024]
The embodiment of the template setting method has been described above with an example. In the above description, the horizontal differentiation process (S1), the vertical differentiation process (S6), and the binarization (S2, S7) are examples of the edge extraction process (claims). The vertical reduction process (S3), horizontal reduction process (S8), edge pixel number storage (S4, S9), address string creation (S5, S10), template address setting (S11) are the region selection process (claims). It is an example.
[0025]
Next, a modification example (1) of the above-described embodiment will be described. In the position of the template selected in the template address setting (S11) described above, when a similar pattern exists in the misalignment correction area, the degree of coincidence becomes maximum in the vicinity of the similar pattern and misplaces the real pattern. There is a risk of calculating the quantity. Therefore, in the template address setting (S11), an area having no similar pattern in the misalignment correction area is selected as a template.
[0026]
Therefore, a process for checking the presence or absence of a similar pattern is performed. The template selected in the template address setting (S11) described above is applied, and the template matching is performed by relatively changing the positions of the template and the reference image within the misalignment correction region. When a similar pattern does not exist in the misalignment correction area, the degree of coincidence becomes maximum only at one place. For example, if the evaluation value of the degree of coincidence is a correlation coefficient, the correlation coefficient is a maximum value and a maximum value at the template position (reference position; real picture position) as it is, and is naturally “1”.
[0027]
When similar patterns exist in the misalignment correction area, the degree of coincidence becomes maximum at a plurality of locations. For example, when the evaluation value of the degree of coincidence is a correlation coefficient, the maximum value is smaller than “1” at the position of the similar pattern. Therefore, when a predetermined threshold is set and the degree of coincidence is maximal at a plurality of locations, and the degree of coincidence is higher than the predetermined threshold, the template is inappropriate and excluded from the template to be actually used. Perform the process. For example, if there are two or more maximum values having a correlation coefficient of “0.7” or more, the template is excluded. This threshold value is set to be a value smaller than the lower limit of the correlation coefficient when the inspection object is a regular one (non-defective product).
[0028]
For the template area selected in the template address setting (S11) described above, the above-described processing for checking the presence or absence of a similar pattern is performed. If a similar pattern exists, the template is excluded, and the above-described template address setting (S11) is further continued. When a similar pattern does not exist in the selected template area, the template is set as a template to be actually used.
[0029]
Next, a modified example (No. 2) of the above-described embodiment will be described. If the template size can be reduced at the template position selected in the template address setting (S11) described above, the amount of calculation is reduced in the calculation of the degree of coincidence, so that the inspection speed can be increased. That is, if an area where edges in a predetermined direction are closely present is selected, the template size can be reduced. Therefore, in the template address setting (S11), the region where the edges in the predetermined direction are connected is selected as a template.
[0030]
Therefore, a process for checking whether or not the edge in the predetermined direction is a connected region is performed. The template selected in the template address setting (S11) described above is applied, and it is determined whether or not there is a common pixel for the vertical edge image and the horizontal edge image in the template area. For example, the pixel value of the pattern (edge) portion is “1”, and the pixel value of the background portion is “0”. At that time, if the logical operation “AND” is performed for each corresponding pixel in the vertical edge image and the horizontal edge image. An edge common partial image can be obtained as an image of common pixels.
[0031]
In the edge common portion image, when there is a portion having a pixel value “1” indicating the common portion of the edge, the vertical edge image and the horizontal edge image are connected. Note that the determination of whether or not they are connected is not limited to the method described above. For example, the determination may be made by calculating the connectivity number in binary image processing.
[0032]
The above-described processing for checking whether or not the template area selected in the template address setting (S11) is linked is performed. If not connected, the template is excluded, and the above-described template address setting (S11) is further continued. Then, when the vertical edge image and the horizontal edge image are connected in the selected template region, the template is used as a template prototype to be actually used.
[0033]
Then, a process for reducing the area of the original template to be actually used is performed. For example, in the edge common portion image, the template region is reduced as a region having a predetermined size that wraps around a portion having a pixel value “1” indicating the common portion of the edge. The reduced template is set as a template to be actually used.
[0034]
【The invention's effect】
As described above, according to the template setting method according to claim 1 of the present invention, an appropriate template capable of obtaining high accuracy and stability is set as a template for calculating the positional deviation amount. .
Further, according to the template setting method of the second aspect of the present invention, the calculation amount is reduced in the calculation of the degree of coincidence, so that the inspection speed can be increased.
Further, according to the template setting method of the third aspect of the present invention, an appropriate template that can obtain particularly high precision and stability is set.
Further, according to the template setting method according to claim 4 of the present invention, it is possible to avoid an error in miscalculating a misalignment between a similar pattern and a real pattern.
[Brief description of the drawings]
FIG. 1 is a diagram showing a data processing process in a template setting method of the present invention.

Claims (4)

An edge extraction process for extracting edges in a reference image obtained by imaging a printed matter as an inspection reference, and counting the number of edge pixels in a template size region at a predetermined position with respect to edges in at least two predetermined directions; Create a position sequence in which the positions of the template size areas are rearranged in the order of the number of edge pixels in each edge direction, and the template size in the order of the number of edge pixels in either direction in the position sequence in each edge direction. And a region selection step of selecting the region of the reference image at the position of the region of the common template size that appears first as a template.   2. The template setting method according to claim 1, wherein the region selecting step is a step of selecting, as a template, a region where edges in the predetermined direction are connected.   3. The template setting method according to claim 1, wherein the region selecting step is a step of selecting, as a template, a region of the reference image including many edges in the predetermined direction after removing edges other than the predetermined direction. A template setting method characterized by that.   4. The template setting method according to claim 1, wherein the region selecting step is a step of selecting, as a template, a region having no similar pattern in the misalignment correction region.

Images