JP2002008029A - Image inspection device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image inspection device capable of absorbing a positional slippage error or digital error without deteriorating the resolution. SOLUTION: Each inspecting picture element of an image to be inspected is compared for density difference with a plurality of reference picture elements within a picture element range of, for example, 3×3 about the corresponding reference picture element of a reference image located in a position corresponding to the inspecting picture element. When the smallest value of the density differences of the inspecting element with the 3×3 reference picture elements is smaller than a threshold, the matching of the inspecting element to the corresponding reference picture element is judged.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image inspection apparatus used for an apparatus for inspecting a defect of an inspection object such as a printed picture or the like.

[0002]

2. Description of the Related Art Various devices are known as devices for inspecting a defect such as a printed pattern on an object to be inspected. An apparatus using a differential / differential contour extraction method and a pattern matching method are mainly used. It can be broadly classified into the equipment used. In the differential / differential contour extraction method, a contour image whose density varies greatly spatially is extracted by spatially differentiating or subtracting the reference image, mask data of the contour image is generated, and the differential image is similarly differentiated in the inspection image. Alternatively, the difference is extracted to extract the contour. If there is a defect in the inspection image, an abnormality can be determined because no mask is generated for the defect. In one pattern matching method, a reference image is created and stored, and each reference pixel of the stored reference image is compared with each inspection pixel of the inspection image for each pixel to determine matching.

Conventionally, defect detection by the differential / differential contour extraction method has been generally used. However, this method is difficult to detect a thin stain or a defect whose color gradually changes over several patterns. There are disadvantages. For thin stains, the differential waveform does not change sharply, so it does not reach the detection level.
Basically, since this method only sees a change in the same image pattern, it is difficult to make a difference in density, and it is difficult to detect a defect. On the other hand, the pattern matching method has an advantage that it has a high detection capability for a light stain or a defect whose color changes gradually because the density difference of each pixel between the reference image and the inspection image is determined.

[0004]

However, in the conventional pattern matching method, when there is a position error or a digital error in spatial sampling,
Even though the inspection image does not actually have a defect, there is a risk that the inspection image is erroneously detected as having a defect because it does not match the reference image.

In order to avoid such erroneous detection, it is conceivable to perform a smoothing process for averaging neighboring pixels. In particular, there is a problem that the detection accuracy of linear or point defects is deteriorated.

SUMMARY OF THE INVENTION It is an object of the present invention to provide an image inspection apparatus capable of absorbing a position error and a digital error without lowering the detection accuracy.

[0007]

In order to achieve the above object, the present invention provides an image pickup apparatus for picking up an object to be inspected, an image generating circuit for generating an inspection image from an output of the image pickup apparatus, An image inspection apparatus comprising: each inspection pixel; and a matching circuit that compares the reference pixel of the stored reference image to determine matching between the inspection image and the reference image. The test pixel is compared with the density difference between a plurality of reference pixels within a predetermined range around the corresponding reference pixel of the reference image at the position corresponding to the test pixel, and includes the test pixel and the corresponding reference pixel. If the value of the smallest density difference among the plurality of reference pixels centered on the corresponding reference pixel is smaller than a predetermined threshold, it is determined that the inspection pixel matches the corresponding reference pixel. Do The features.

A density difference between the inspection pixel and the plurality of reference pixels centered on the corresponding reference pixel including the corresponding reference pixel is obtained, and matching is determined using a value having the smallest density difference. Even if a digital error exists, the error can be absorbed.

[0009] Optionally, based on the reference pixel, a contour portion in which the density is spatially largely changed, an expanded portion close to the contour portion, and a contour portion divided into a contour portion and a base portion other than the expanded portion. It is preferable that the apparatus further comprises a section detection circuit, and a threshold variable circuit that changes the predetermined threshold according to whether the corresponding reference pixel belongs to a contour section, an expansion section, or a base section. By setting the threshold value to each characteristic of the contour portion, the inflated portion, or the base portion, erroneous detection can be reduced, and appropriate determination can be made. For example, in the contour portion where the density change is large, the judgment is loose, that is, the threshold is large. That is, the threshold value can be set to be smaller.

[0010] Optionally, the light receiving element of the image pickup device is a color image sensor, and the matching between the inspection pixel and the corresponding reference pixel can be determined for each color component. Thus, matching for each color component can be determined without losing color information. Additionally, optionally,
The outline detection circuit may extract an outline for each color component in which the density of the reference pixel greatly changes for each color component, and may use the outline for each color component as the outline. Since the outline may be different for each color component, by extracting the outline part for each color component, it is possible to extract an outline that can be visually confirmed that is close to human visual sensitivity.

[0011] Optionally, said contour detection circuit comprises:
In addition to detecting a reference image contour portion in which the density greatly changes spatially in the reference image and an inspection image contour portion in which the density greatly changes spatially in the test image, the reference image contour portion and the test image contour portion are respectively detected. The result of the enlargement process and its logical product can be used as the contour portion. By taking the logical product of both the reference image contour part of the reference image and the inspection image contour part of the inspection image (however, by enlarging them in consideration of positional deviation) as the contour part, the inspection image is obtained. In the case where there is a defect such as a chip, matching can be determined based on a strict threshold value other than the contour portion instead of the threshold value of the contour portion. Accordingly, when the reference image has many contours and a defect such as a chip in the contour of the inspection image is to be determined, the defect such as the chip and an error in a normal range in the contour are identified. be able to.

[0012]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram of an entire defect inspection apparatus including an image inspection apparatus according to the present invention. The image inspection apparatus 10 inspects a printed sheet S (inspection object) flowing by the rolls R, R..., And roughly includes an image pickup apparatus 12, an image generation circuit 14, and a defect detection circuit 16 as a matching circuit. And

The printed material sheet S can be, for example, a printed material in which the same picture (pattern) is repeated at a constant cycle, and is run at a substantially constant speed by the rolls R, R. A three-line color CCD camera 12 as an image pickup device is fixed above the traveling path of the printed sheet S. A plurality of cameras 12 can be juxtaposed according to the width of the printed sheet S as appropriate. As an illumination device for the camera 12, a reflection illumination unit 18 is provided on the side of the camera 12 with respect to the printed sheet S, and a transmission illumination unit 20 is provided on the side opposite to the camera 12. The output from the camera 12 is sent to an image generation circuit 14, where an inspection image is generated, and the inspection image is sent to a defect detection circuit 16.

The output from the defect detection circuit 16 is sent to a computer 24 via an interface 22 to be subjected to software processing, and the defect image is stored in a defect image storage unit 26 as storage means. Alternatively, the image is printed by a color printer 28 as a printing unit, and further displayed on a CRT monitor 30 as a display unit.

An encoder unit 32 for detecting a position of the printed material sheet S in the flow direction (also referred to as a Y direction) is provided on a traveling path of the printed material sheet S, and an output of the encoder unit 32 is supplied to the image generating circuit. It is sent to 14.

FIG. 2 is a detailed block diagram of the image generation circuit 14. The output from the camera 12 is individually captured for each of R, G, and B, and is processed in parallel by the R, G, and B modules. However, since each module has the same configuration, one module is representative. To represent. That is, the image generation circuit 14 A / D-converts the output of the camera 12 and outputs the digital density gradation.
Since the position of the imaging line of the camera 12 differs depending on the color, a line correction circuit 44 for correcting the difference is provided.
And meandering correction circuit 4 for correcting meandering of printed sheet S
6, a Y direction cutout circuit 48 for cutting out the flow direction by a pattern trigger based on an output from the encoder unit 32, and an input frame memory 50 for storing an inspection image. The input frame memory 50 is F
It has an IFO structure, and its data is sequentially sent to the defect detection circuit 16 according to the extraction timing from the Y-direction extraction circuit 48.

FIG. 3 is a block diagram showing a detailed structure of the defect detection circuit 16. As shown in FIG. The defect detection circuit 16 mainly includes a reference image memory 60 (color components R (red), B (blue), and G) that stores the inspection image from the image generation circuit 14 as a reference image.
(Has reference image memories 60A, 60B, 60C for each (green)), 61 (has reference image memories 61A, 61B, 61C for each of the color components R (red), B (blue), and G (green))
And a contour detection circuit 62 that separates the reference image from either the reference image memory 60 or 61 into a contour, an inflated portion, and a base, a threshold variable circuit 64, and an inspection image and a reference image. The comparison circuit 66 includes a comparison circuit 66 that determines a pixel determined to be unmatched as a defective pixel, and a defect counter 68 that counts the number of defective pixels based on the defective pixels from the comparison circuit 66 and detects the coordinates. The number of defective pixels and their coordinates from the defect counter 68 are sent to the computer 24 via the interface 22. In the figure, D is a one-line delay circuit (the same applies to other figures).

For the reference image, a method of preparing the data of the reference image in advance as an absolute image, and always using the absolute image as the reference image, and comparing the absolute image with the inspection image can be adopted, or the reference image can be updated under certain conditions. In this embodiment, the inspection image captured by the image generation circuit 14 is stored in the reference image memories 60 and 61, and the inspection image immediately before the inspection image to be inspected is adopted as the reference image. The reference image is updated every time. However, when the inspection image includes a defect, the reference image memories 60 and 61 for storing two inspection images are provided so that the inspection image is not used as the reference image. , 61 are used. This reference image data is supplied to the next contour detection circuit 62 and comparison circuit 66.
Sent to.

In detail, the contour detecting circuit 62 includes a contour detecting section 72, an OR gate 74, and an expanded section detecting section 76. The contour detection unit 72 calculates the R component, the B component,
It comprises contour detection units 72A, 72B, 72C for each color component for detecting a contour for each G component. The edge is emphasized by multiplying a pixel within a 3 × 3 range around the pixel of interest by a predetermined coefficient, and when the value is larger than a predetermined threshold, the pixel of interest is a color component. Output as belonging to each contour part. Accordingly, when it is determined that the target pixel belongs to any of the R, B, and G components, the next OR gate 74 is determined.
Then, by performing a logical sum of the outputs from the respective contour part detection units 72A, 72B, and 72C, it is determined that the pixel of interest belongs to the contour part. The logical sum of the contours for each color component is calculated in this way because the contours may be different for each color component, and even if a contour can be visually confirmed, a certain color component may not be determined to be a contour. It is. O
By calculating the logical sum in the R gate 74, it is possible to stably extract a contour close to human visual sensitivity.

Next, a pixel existing inside or outside by several pixels from the contour portion is defined as an expanded portion by an expanded portion detection portion 76.
Further, a signal is output in accordance with a pixel of interest, using a portion other than the contour portion and the dilated portion as a base portion. In this way, the contour detection circuit 62 classifies the pixel of interest into a contour, an expanded portion, and a base. FIGS. 4A and 4B show an example thereof, in which FIG. 4A shows a reference image, and FIG.
4. Several pixels inside and outside the contour portion 84 (2 in this example)
This is an image divided into an inflated portion 86 and a base portion 88 other than the inflated portion 86 existing in the pixel.

The pixel of interest is a contour portion 84, an expanded portion 86,
When a signal indicating that the signal belongs to any of the base sections 88 is sent from the contour detection circuit 62 to the threshold variable circuit 64, the threshold variable circuit 64 sets a different threshold value according to the signal to a comparison circuit described below. 66 is output. Three types of thresholds are set in advance for the contour portion, the inflated portion, and the base portion before the inspection. The threshold value Th1 is large so that the contour portion 84 can be a soft decision, the threshold value Th2 is small so that the base portion 88 can be a strict decision, and the expansion portion 86 can tolerate a position error and a digital error of the contour portion 84. The threshold value Th3 is set to an intermediate value (Th2 <Th3 <Th1).

Next, in the comparison circuit 66, the image generation circuit 1
A comparison is made between the test pixel from No. 4 and the reference pixel from either of the reference image memories 60 or 61. As shown in FIG. 5, a comparison is made between a test pixel of interest and 3 × 3 = 9 reference pixels centering on the corresponding reference pixel. A 1: 9 comparison is performed to determine each density difference. When the smallest value among the nine density differences is smaller than the threshold value, it is determined that the inspection pixel matches the corresponding reference pixel, and conversely, When the value with the smallest density difference is larger than the threshold, it is determined that the inspection pixel and the corresponding reference pixel do not match, and that the inspection pixel is a defective pixel. As the threshold at this time, any of the thresholds Th1, Th3, and Th2 output from the threshold variable circuit 64 depends on whether the corresponding reference pixel belongs to the contour portion 84, the expansion portion 86, and the base portion 88 as described above. used. This 1: 9 comparison is performed for each of the R, B, and G components. Then, if any of the R, B, and G components is determined to be a defective pixel, the signal is output to the next defect counter 68. By performing the 1: 9 comparison, when a positional shift occurs between the inspection image and the reference image or when there is a digital error due to digitizing the space, such an error is absorbed. Can be.

The defect number counter 68 counts the number of defective pixels. If the number of defective pixels is larger than a predetermined number, it is determined that the portion of the printed sheet S corresponding to the inspection image is defective. . Then, the coordinates of the defective pixel are obtained, and the number of defects and the coordinate data are sent to the computer 24 via the interface 22. In the computer 24, if necessary, the image of the defective portion is
The defect number is displayed on the RT monitor 30 and the defect number and coordinate data are stored in the defect image storage unit 26.

FIG. 6 is a detailed block diagram of the contour detection circuit 62 of the defect detection circuit 16 showing another embodiment according to the present invention. The other points are the same as in the first embodiment, and are not shown.

In this embodiment, the outline detection circuit 62 of the defect detection circuit 16 further includes an inspection image outline detection unit 78 for obtaining an outline for the inspection image, and an OR gate 79.
And enlargement processing units 75 and 80 and an AND gate 81. For example, when many characters are included in the printing of the printed sheet, the contour detection circuit 62 determines many parts as contours.
The threshold value output from 4 becomes weak, and even if a part of the character in the inspection image is missing or missing, such a defect may not be detected. Therefore, in the present embodiment, the contour is not determined only by the reference image, but is also detected in the inspection image, and only when the contour is detected in both the reference image and the corresponding area of the inspection image. , The threshold value for determining the defect detection as a correct contour portion is moderated.

As in the case of the contour detecting section 72, the inspection image contour detecting section 78 detects contours for each of the R, B, and G components.
B, 78C, and each of the contour detection units 78A, 78B,
The 78C has a Sobel operator, for example, performs differentiation processing, and performs 3 × centering on a pixel of interest in the inspection image.
An edge is emphasized by multiplying a pixel within the range of 3 by a predetermined coefficient, and when the value is larger than a predetermined threshold value, the pixel of interest is output as belonging to the contour portion. The ORs of the contours extracted for each of the three components are calculated by an OR gate 79.

The enlargement processing units 75 and 80 are provided with a contour detection unit 7
2 and enlarges the outline detected by the inspection image outline detection unit 78. Specifically, the outline is expanded by several pixels inward and outward. This enlargement processing is performed when the position of the reference image and the inspection image is displaced.
The logical product of the D-gate 81 prevents the output of nothing even though the contours exist in both images, and expands according to the assumed positional deviation.
Next, the AND gate 81 calculates the logical product of the enlarged outlines from the enlargement processing units 75 and 80, and sets the logical product as the outline 84. FIG. 7 shows an example of the processing, and (a)
-1) represents an inspection image, and (a-2) represents a reference image. A defect 90 exists in the inspection image, and a position shift exists in the X direction between the inspection image and the reference image. (B
-1) and (b-2) are contours extracted in the inspection image and the reference image, respectively, and (c-1) and (c-2) are (b-1) and (b-2), respectively. (D) show the results of (c-1) and (c-
This is a contour portion 84 obtained by taking a logical product with 2). (D)
As can be seen from the figure, since the portion corresponding to the lack 90 of the inspection image is not discriminated from the contour portion 84, it is possible to prevent all the threshold values selected by the threshold variable circuit 64 from becoming loose. Can be.

[0028]

As described above, according to the present invention,
The density difference between each test pixel of the test image and a plurality of reference pixels within a predetermined range around the corresponding reference pixel of the reference image at the position corresponding to the test pixel is compared, and the test pixel is compared with the test pixel. When the value of the smallest density difference among the plurality of reference pixels centered on the corresponding reference pixel including the reference pixel is smaller than the threshold value, the inspection pixel and the corresponding reference pixel are matched. Since it is determined that there is an error, it is possible to absorb a position error and a digital error without lowering the resolution.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating an embodiment of an image inspection apparatus according to the present invention.

FIG. 2 is a detailed block diagram of the image generation circuit of FIG. 1;

FIG. 3 is a detailed block diagram of the defect detection circuit of FIG. 1;

FIG. 4A is an example of a reference image, and FIG. 4B is an example of an image showing a contour portion, an expanded portion, and a base portion with respect to FIG.

FIG. 5 is an image example showing a 1: 9 comparison.

FIG. 6 is a detailed block diagram of a contour detection circuit of a defect detection circuit showing another embodiment according to the present invention.

FIG. 7 is an image example for understanding the operation of the contour detection circuit of FIG. 6;

[Explanation of symbols]

 12 Camera (imaging device) 14 Image generation circuit 16 Defect detection circuit (matching circuit)

Continued on the front page (72) Inventor Mikiya Tanaka 2-16-46 Minami Kamata, Ota-ku, Tokyo Inside Tokimec Co., Ltd. F term (reference) 2F065 AA56 BB02 BB15 CC02 FF04 JJ03 JJ26 MM03 QQ08 QQ13 QQ24 QQ29 QQ32 QQ39 SS06 SS13 2G051 AA34 AB02 AB11 AC21 CA03 CA04 CB01 DA06 EA12 EA14 EA17 DB02 DA07 DB12 DA07 DB07 DC33 5L096 AA02 AA06 BA03 BA20 CA02 DA02 EA02 FA06 GA08 GA22 GA23 GA51 HA08

Claims (1)

[Claims] 1. An imaging device for imaging an object to be inspected, an image generation circuit for generating an inspection image from an output of the imaging device, each inspection pixel of the inspection image, and a reference pixel of a stored reference image. A matching circuit for comparing and matching between the inspection image and the reference image, wherein the matching circuit includes: each inspection pixel of the inspection image; and a reference image at a position corresponding to the inspection pixel. Each of the density differences between a plurality of reference pixels within a predetermined range around the corresponding reference pixel is compared, and the test pixel and the plurality of reference pixels around the corresponding reference pixel including the corresponding reference pixel are compared. An image inspection apparatus characterized in that when the value of the smallest density difference among the density differences is smaller than a predetermined threshold value, it is determined that the inspection pixel matches the corresponding reference pixel.