JPH1139487A - Method and device for image inspection - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make a quality decision wherein the difference of a defect from a normal density value and the size of area are both integrated by binarizing a morphology operation image obtained from a light-shade image through operation with a specific threshold value. SOLUTION: When an image input camera 1 picks up an image of a print surface in synchronism with conveyance and stores it in an input buffer 2, a difference and absolute value generating circuit 8 inputs an image to be inspected from the buffer 2, a reference image from a reference image memory 4, and a blind sector image from a blind sector image memory 6 to obtain an absolute value image as the difference between the reference image and object image. Morphology operation circuits 9a to 9c generate a morphology image for the absolute value image of the difference by using a specific constitution function. Then binarizing circuits 11a to 11c binarize the morphology operation image according to a threshold value. Consequently, a defective area is discriminated according to the difference from the normal density value and the size of the area, a decision circuit 13 outputs a defect kind value, and a control output circuit 14 performs a process corresponding to the value.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image inspection method and apparatus for comparing and inspecting a reference image and an image to be inspected.
In particular, the present invention relates to an image inspection method and apparatus capable of detecting a defect having a large difference from a normal density value but a small area and a defect having a large area but a small difference from a normal density value.

[0002] When the density of printing on a printed matter is quantitatively represented, "density value" is generally used. But,
The density of the print is not limited to the “density value” and can be represented by “reflectance”, “transmittance”, or other linear or non-linear quantity. In the present invention, the density of printing expressed quantitatively is described as "density value" in a sense including other numerical expressions such as "reflectance".

[0003]

2. Description of the Related Art There are various defects in printed matter, such as defects such as wrinkles, foreign matters, and stains on printing paper, and ink stains, hickey, doctor streak, poor density, and poor color tone that occur during printing. is there. Conventionally, printed matter inspection is mainly performed by visual inspection performed by humans. For example, in the case of printed matter discharged from a folding unit of a sheet-fed printing press or a rotary printing press, a printing machine operator (operator) ) Was removed at the appropriate time and inspected visually.
If a continuous printout is ejected from the printing press and cannot be extracted, the printout is irradiated with flash light that flashes instantaneously in synchronization with the running speed of the printout, and the printout is made stationary using the afterimage of the human eye. Was visually inspected.
Visual inspection requires concentration of the mind and requires a very large work load on the person to be inspected. Further, since the inspection is performed by a person, the inspection standard is ambiguous and it is inevitable to miss a defect. Therefore, proposals have been made to automatically perform printed matter inspection, and various techniques have been disclosed.

For example, in Japanese Patent Publication No. 1-47823, image data read from a picture of a running printed matter when a normal printed matter is printed is stored as reference image data in an image memory, and read out from the image memory. There is disclosed a technique relating to a printed matter inspection apparatus of a method of comparing the reference image data read from a pattern of a printed matter to be inspected during printing with the image data to be inspected on a pixel-by-pixel basis to determine the quality of printing.

[0005]

However, since such an inspection apparatus is based on comparison on a pixel-by-pixel basis, a defect having a large difference from a normal density value can be obtained regardless of the size of the area. There is a problem that a defect that can be detected but has a small difference from the normal density value cannot be detected no matter how large its area is. In order to avoid this, by integrating the absolute value of the difference from the normal density value of the pixel in the predetermined area, it is possible to detect the difference from the small normal density value, which is difficult to distinguish from noise, as a clear significant difference. Done. Since it is easy to realize the integration processing device, the present invention is applied to detection of a slight change in the density value of the entire print image and detection of a doctor streak or a hitkey extending in the transport direction of the printing paper.

However, originally, it is effective to determine the area to be integrated in accordance with the content of the printed picture. However, since the patterns in printed matter are various, it is not practical for operation to determine the area to be integrated each time the print item is switched. Further, it is conceivable to diversify and subdivide the region to be integrated in advance so as to be able to cope with various kinds of pictures, but this is not realistic in terms of the scale of the apparatus because the amount of calculation in the integration processing apparatus is enormous. SUMMARY OF THE INVENTION It is an object of the present invention to provide an image inspection method and apparatus for performing a pass / fail judgment based on an inspection criterion that treats both a difference between a normal density value in a defect and an area size as one.

[0007]

The above objects are achieved by the present invention described below. That is, the present invention provides an image inspection method for comparing and inspecting a reference image and an image to be inspected, wherein a morphological operation is performed on a grayscale image using a predetermined configuration function to obtain a morphological operation image; An image including a binarization process of performing a binarization operation on the morphology operation image using the threshold value to obtain a binarized image, and a pass / fail determination process of performing pass / fail judgment based on the binarized image Inspection method ".
According to the present invention, the morphology operation image obtained in the morphology operation process is configured by extracting a partial image in a specific range with respect to the pixel value (density value) and the number of pixels (area) from the grayscale image according to the configuration function. It can be an image. In the subsequent binarization process, the binarized image can be an image indicating whether or not it is included in a specific range defined by the pixel value (density value) according to the threshold value. Then, in the pass / fail judgment process, 2 obtained above is obtained.
Since the pass / fail judgment is performed based on the digitized image, an image inspection method is provided in which the pass / fail judgment is performed based on an inspection criterion that treats both the difference from the normal density value and the area size of the defect as one.

[0008] The present invention also provides that "the morphological operation step is a step of performing a morphological operation on a grayscale image using a plurality of predetermined constituent functions to obtain a morphological operation image corresponding to each of the constituent functions. The binarization process performs a binarization operation on a morphological operation image corresponding to each of the constituent functions using a predetermined threshold value corresponding to each of the constituent functions, and performs a binary operation corresponding to each of the constituent functions. Image inspection method, wherein the pass / fail determination step is a step of performing pass / fail determination based on all of the binarized images corresponding to each of the constituent functions. "
It is. According to the present invention, the morphology operation image obtained in the morphology operation process is configured by extracting a partial image in a specific range with respect to the pixel value (density value) and the number of pixels (area) from the grayscale image according to the configuration function. It can be an image. In the subsequent binarization process, the binarized image can be an image indicating whether or not it is included in a specific range defined by the pixel value (density value). A plurality of binarized images are obtained by a plurality of constituent functions and a plurality of thresholds corresponding thereto. In the pass / fail judgment process, pass / fail judgment is performed based on the plurality of binarized images obtained as described above. Therefore, pass / fail judgment is performed based on an inspection criterion that treats both the difference between the normal density value and the area size of the defect as one. Further, an image inspection method is provided which has a high degree of coincidence between the inspection standard and a desired standard.

Further, the present invention is an image inspection method wherein the grayscale image is an image obtained by calculating at least an absolute value of a difference based on a reference image and an image to be inspected. According to the present invention, since the pixel value (density value) and the number of pixels (area) of the grayscale image are directly linked to the difference between the normal density value and the area size of the defective image, the morphological calculation process is simplified. In addition, the present invention provides a method of determining whether a defect type is based on both defect attributes of a difference between a normal density value and a size of a defect based on all binarized images corresponding to each of the constituent functions. Image inspection method for determining and obtaining defect type values "
It is. According to the present invention, it is possible to determine whether a defect detected by the defect type value is a heavy defect or a light defect. The present invention is also an "image inspection method including a control step of outputting an operation amount corresponding to the defect type value". According to the present invention, it is possible to automatically perform an appropriate operation such as an alarm or a print stop according to a defect type value by a control process. Further, the present invention is "an image inspection method in which the morphological operation is an opening operation". According to the present invention, a morphological operation image formed by extracting a partial image in a specific range from a gray image with respect to a pixel value (density value) and the number of pixels (area) in accordance with the above-described configuration function is obtained by a simple operation. It can be obtained in a compatible form. Further, the present invention is an "image inspection method in which a predetermined threshold value corresponding to each of the constituent functions corresponds to a small threshold value for a constituent function having a large area and / or a density value". According to the present invention, the general recognition of the weight of a defect matches the tendency of a detected defect.

Further, according to the present invention, there is provided an image inspection apparatus for inspecting a reference image by comparing a reference image with an image to be inspected. Morphological operation means for obtaining a morphological operation image to perform, and performing a binarization operation on the morphological operation image corresponding to each of the constituent functions using a predetermined threshold value corresponding to each of the constituent functions, and performing each of the constituent functions A binarization unit that obtains a binarized image corresponding to, and a pass / fail determination unit that performs pass / fail determination based on all of the binarized images corresponding to each of the constituent functions.
Image inspection apparatus having According to the present invention, the morphology operation image obtained by the morphology operation means is configured by extracting a partial image in a specific range with respect to the pixel value (density value) and the number of pixels (area) according to the configuration function from the grayscale image. It can be an image. In the subsequent binarizing means, the binarized image can be an image indicating whether or not it is included in a specific range defined by the pixel value (density value). A plurality of binarized images are obtained by a plurality of constituent functions and a plurality of thresholds corresponding thereto. Then, in the pass / fail determination means, a plurality of 2
Since the pass / fail judgment is performed based on the binarized image, the pass / fail judgment is performed based on an inspection criterion that treats both the difference from the normal density value in the defect and the size of the area as a unit. An image inspection device having a high degree of coincidence is provided.

[0011]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Next, the present invention will be described with reference to embodiments. FIG. 1 is a block diagram showing the configuration of an apparatus (image inspection apparatus) for implementing the image inspection method of the present invention.
In FIG. 1, reference numeral 1 denotes an image input camera for capturing an image of a printing surface;
2 is an image input buffer for temporarily storing an image, 3 is a synchronization circuit for synchronizing image input, 4 is a reference image memory for storing an image of a non-defective printed matter (reference image), and 5 is a dead zone image calculated from the reference image. A dead zone calculation circuit, 6 is a dead zone image memory for storing a dead zone image, 7 is a circuit for instructing updating of a reference image and creation of a dead zone image, and a reference image updating / dead zone image creation command circuit, and 8 is a circuit for controlling an image to be inspected and a reference image. A difference / absolute value conversion circuit for calculating the absolute value of the difference, 9a,
Reference numerals 9b and 9c denote morphological operation circuits for performing morphological operations on images (absolute difference value images). Reference numeral 10 denotes a configuration function setting circuit for setting operation parameters (configuration functions and the like) in the morphological operation circuits 9a, 9b and 9c.
11b and 11c are binarization circuits for binarizing an image (morphologically processed image), and 12 is a binarization circuit 11a, 1
Set the calculation parameters (threshold and the like) of 1b and 11c 2
A binarization threshold setting circuit, 13 is a binarization circuit 11a, 11b,
A determination circuit 14 determines the quality of the printed matter based on the output of 11c, and a control circuit 14 outputs an operation amount corresponding to the failure type value output by the determination circuit 13.

The image input camera 1 is installed in a printing machine and captures an image of a printing surface of a web (printing paper) discharged from a printing unit. The image input camera 1 is a line sensor camera,
An image sensor camera or the like can be used. An image signal (analog signal) from the image input camera 1 is converted into digital data by an A / D converter, and
Is temporarily stored. When the inspection is performed, the web is transported in the printing press. Therefore, in synchronization with the transport of the web, the image input camera 1 captures an image of a printing surface, and the image input buffer 2 temporarily stores an image of a predetermined print area. . This synchronization is performed based on a synchronization signal. The synchronizing circuit 3 generates a synchronizing signal based on a web transfer signal output from, for example, a rotary encoder that detects rotation of a driving shaft of a printing press, and outputs the synchronizing signal to the image input camera 1 and the input buffer 2. .

The reference image memory 4 receives an image temporarily stored in the image input buffer 2 and stores an image of the entire predetermined printing area, that is, an image of a predetermined inspection area (an image of a printed matter for one cycle). This image stored in the reference image memory 4 is an image of a non-defective printed matter (reference image). When a non-defective product is being printed on the printing press, the circuit reference image update / dead zone image creation command circuit 7 stores the reference image memory 4 in the reference image memory 4 based on logical judgment automatically performed by the control panel of the printing press or manual input by an operator. An output for instructing the updating of the reference image is performed. The updating of the reference image is performed not only before the start of the inspection, but also during the inspection, for example, when the printing conditions are changed, for example, when the printing speed is changed. FIG. 2A shows an example of the reference image. Although the reference image is originally a grayscale image (in this example, the pixel value is 0 to 255), the grayscale image is shown by a line drawing in FIG.

The dead zone calculation circuit 5 calculates a dead zone image based on the reference image stored in the reference image memory 4 and outputs the calculated dead zone image to the dead zone image memory 6. When comparing the reference image and the inspection target image for each pixel, generally, the edge portion (the portion where the density changes rapidly) of the image is greatly affected by the positional deviation between the reference image and the inspection target image, and the positional deviation is within a normal range. Also, the absolute value of the difference becomes large. The dead zone image is data for preventing such misregistration within a normal range from being erroneously determined as a printing failure. That is, when evaluating the difference between the reference image and the inspection target image,
Rather than evaluating only by the absolute value of a simple difference, a value set in a dead zone image is considered. For example, when the absolute value of the difference exceeds a value set in the dead zone image, the exceeded value is regarded as a substantial difference between the reference image and the inspection target image.

The dead zone calculation circuit 5 calculates a dead zone image based on a command output from the circuit reference image update / dead zone image creation command circuit 7. Normally, a dead zone image is calculated each time the reference image is updated. Figure 2 shows an example of the dead zone image
It is shown in (B). The dead zone image is originally a grayscale image,
In FIG. 2B, as in FIG. 2A, a gray image is shown by a line drawing. In FIG. 2B, the dark portion indicates the size of the dead zone (the intensity of the dead zone).

The difference / absolute value conversion circuit 8 inputs the inspection target image output from the image input buffer 2, the reference image stored in the reference image memory 4, and the dead zone image stored in the dead zone image memory 6. Then, the absolute value of the difference between the reference image and the inspection target image is calculated for each pixel to obtain an absolute value image of the difference. As described above, the dead zone image is considered in the calculation. For example, when the absolute value of the difference between the inspection target image and the inspection target image for each pixel exceeds the value set in the dead zone image, the exceeded value is regarded as the absolute value of the difference. The value of each pixel of the value image is used. If the value does not exceed the value set in the dead zone image, there is no difference, and the pixel value is set to “0”.

FIG. 2 (C) shows an example of an image to be inspected, FIG. 2 (D) shows an example of a simple difference absolute value image not considering a dead zone image, and FIG. FIG. 3E shows an example of the absolute value image. Although these are originally grayscale images, FIGS. 2C, 2D and 3
In (E), a grayscale image is shown by a line drawing. The inspection target image shown in FIG. 2C includes a defective portion. FIG. 2 (D)
In the absolute value image of the simple difference shown in (1), an edge portion (not a defect) in the image is recognized. In the absolute value image of the substantial difference shown in FIG. 3E, the edge portion in the image is almost disappeared and disappears, but a difference between the inspection target image which is not a defective portion and the reference image also appears.

Morphological operation circuits 9a, 9b, 9c
Is the morphology (morpholog
y) Perform the operation. Morphological operations are performed on the target image and the structuring functio
The basic operation is a binary operation with n). Various transformations can be obtained by changing the configuration function. The constituent function setting circuit 10 converts the constituent function into a morphological operation circuit 9
a, 9b, and 9c. The morphological operation is “di
It is composed of four basic operations: “lation”, “erosion”, “opening”, and “closing.” The morphological operations will be described later, and the role of the morphological operation circuits 9a, 9b, and 9c in FIG. .

In order to make it easy to understand, the case where the morphological operation circuits 9a, 9b and 9c in FIG. 1 are omitted and the case where they are present will be compared. When the morphology operation circuits 9a, 9b, 9c are omitted, the binarization circuit 11
Reference numerals a, 11b, and 11c binarize the absolute value image of the difference output from the difference / absolute value conversion circuit 8. The threshold value for performing the binarization is set in the binarization circuits 11a, 11b, and 11c by the binarization threshold setting circuit 12. FIG. 3 (E)
The binarization circuit 1 for the absolute value image of the substantial difference shown in FIG.
When binarization is performed by 1a, 11b, and 11c, FIG.
(F), the binary images shown in FIGS. 3 (G) and 3 (H) are obtained. 3 (F), 3 (G) and 3 (H) shown in FIG.
Two binarized images are obtained by performing binarization at different thresholds.
FIG. 3F shows a binarized image having the smallest threshold value (18).
/ 255; threshold value / maximum value), FIG. 3G is an intermediate threshold value (36/255), and FIG. 3H is the largest threshold value (54/255).

The decision circuit 13 includes a plurality of binarization circuits 11
The pass / fail judgment is made based on the presence / absence of a pixel exceeding a threshold value (the presence / absence of a black pixel) in the binary image output by a, 11b, and 11c. Although there is a method of making a determination in consideration of the continuity and shape of black pixels, the determination is basically made based on the presence or absence of a black pixel. Therefore, when the three binarized images are examined, the following can be understood. In FIG. 3F, it can be seen that a defect is detected and a non-defective part is erroneously detected as a defect (noise is mixed). Further, in FIG. 3G, the degree of erroneously detecting a defect is small, but not all. Moreover, oversight of defects is recognized. Further, in FIG. 3H, there is no erroneous detection of a defect, but it is extremely difficult to overlook the defect. Therefore, when simple binarization is performed on the absolute value image of the difference as shown in this example, erroneous detection cannot be avoided when trying to detect a defect with low density, and erroneous detection is avoided. In this case, a defect having a low density is overlooked.

FIG. 4 is a diagram showing a general tendency of human recognition of defective density (difference from normal density) and defective area. In FIG. 4, the horizontal axis indicates the defective area and the vertical axis indicates the defective density. Generally, the larger the defect density and the larger the defective area, the more the defect is regarded as a heavy defect, and conversely, the smaller the defect density and the smaller the defective area, the lighter the defect. In FIG. 4, the farther away from the origin in the upper right direction, the heavier the defect, and the closer to the origin in the lower left direction, the lighter the defect. Therefore, as shown in FIG. 4, the area can be classified into a heavy defect area, a medium defect area, a light defect area, and a normal area.

Morphological operation circuits 9a, 9b, 9c
Is omitted, the above-described binarization circuits 11a, 11a
Binarization by b and 11c is equivalent to dividing the region in FIG. 4 into two by a straight line parallel to the horizontal axis (defective area) in FIG. Then, a defect included in an area above the straight line is detected, and a defect included in an area below the straight line is not detected. This corresponds to a conventional method of detecting a defect only by a level difference. The area above the straight line includes the above-described normal area, which indicates that erroneous detection of a defect is inevitable even though the area is normal. Also, the area below the straight line includes not only the normal area but also the area of heavy defect, the area of medium defect, and the area of light defect, showing that it is inevitable to overlook particularly low density defects. I have.

Changing the threshold for binarization corresponds to changing the vertical position of a straight line parallel to the horizontal axis (defective area) in FIG. 4, that is, changing the defect level. However, even if such an attempt is made to change the defect level in the example of FIG. 4, erroneous detection or oversight cannot be basically avoided. That is, when simple binarization is performed on the absolute value image of the difference, erroneous detection cannot be avoided when trying to detect a defect with low density, and a defect with low density cannot be avoided when trying to avoid erroneous detection. Will be overlooked.

Next, the case of the present invention in which the morphological operation circuits 9a, 9b and 9c in FIG. 1 exist will be described. Morphological operation circuits 9a and 9b (9c)
In FIG. 3E, a morphology operation called opening is performed on the absolute value image of the substantial difference shown in FIG.

The morphological operation generalized to a gray-scale image is performed by performing an Umbra transform (conversion from multi-valued data to binary data; shading conversion) once to obtain a voxel (0, 1) in a three-dimensional space. After expressing an image as a voxel (three-dimensional pixel), performing a morphological operation on a three-dimensional binary image, inverse truncation (Surface trnsform; 2)
The operation of performing a conversion from a value to a multi-valued image (surface conversion) and returning the image to a grayscale image is an accurate operation. In this case, dilation and erosion are performed simply by using Max and Min values within the component range of the target pixel without giving the component a depth (shade / tone). Therefore, opening and closing were performed similarly.

As a result, FIG.
And the morphology operation image shown in FIG. 5 (J) is generated. The area of the configuration function used in the processing of FIG. 5J is larger than the area of the configuration function used in the processing of FIG. As shown in FIG. 5 (I) and FIG. 5 (J), by performing this morphological operation, an isolated point and an image portion having a small area are removed. Since the area of the configuration function of FIG. 5J is larger than that of FIG. 5I, the area of the function shown in FIG.
(J) has more image portions to be removed.

FIG. 6 is a diagram showing an example of the configuration function. FIG. 6A has an area composed of 1 × 1 pixels.
FIG. 6B has an area composed of 3 × 3 pixels.
(B) has an area composed of 5 × 9 pixels. Pixels shown in black are central pixels, and pixels shown in gray are neighboring pixels. As shown in FIG. 6, the configuration function is also a kind of image (two-dimensional one-valued function), and each pixel of the configuration function has a pixel value. However, here, it has only two values (0, 1). By performing a morphological operation on the image of FIG. 3E using the configuration function of FIG. 6B, the image of FIG. 5I is obtained, and the image of FIG. By performing the morphological operation using the constitutive function, FIG.
The image of (J) is obtained.

In the morphological operation, the central pixel of the configuration function is applied to the target pixel of the image, and the operation is performed within the pixel of the configuration function (the central pixel and the neighboring pixels). The operation is performed on all the target pixels in the region where the morphological operation is performed by changing the target pixel of the image. FIG. 7 is an explanatory diagram of a method for performing this morphological operation. In FIG. 7, gray pixels represent portions where the density value of the original image is relatively large, and white pixels represent portions where the density value of the original image is relatively small. Originally, the target image is a grayscale image having continuous density pixel values, but is simplified in FIG. In FIG. 7, the bold line indicates a configuration function (here, a square of 3 × 3 pixels).

FIG. 8 is a block diagram showing an example of an arithmetic unit for performing an opening operation which is one of the morphological operations. The operation of the opening operation device will be described with reference to FIGS. An image to be subjected to an opening operation is temporarily stored in the image buffer 21. The configuration function control unit 22 controls the execution of the calculation by applying the central pixel of the configuration function to the sequentially selected target pixels. Under the control of the configuration function control unit 22, the minimum value calculation unit 23 calculates the pixel value (density value) of the configuration function and the pixel value (density value) of the calculation target image for each pixel where the configuration function and the calculation target image overlap. ) And the smaller pixel value (density value)
An operation for replacing the pixel value (density value) of the image to be calculated is executed (see FIG. 7). The image buffer 24 temporarily stores an image obtained by the calculation.

The constituent function control unit 25 controls the execution of the calculation by applying the central pixel of the constituent function to the target pixel of the image buffer 24 which is sequentially selected (see FIG. 7). Maximum value calculation unit 26 under the control of configuration function control unit 25
Compares the pixel value (density value) of the configuration function with the pixel value (density value) of the calculation target image for each pixel where the configuration function and the calculation target image overlap, and determines the larger pixel value (density value) ) Is performed to replace the pixel value (density value) of the calculation target image (see FIG. 7). The image buffer 27 temporarily stores an image obtained by the calculation. This allows
The opening operation is performed on the image in the image buffer 21, and the image after the opening operation is temporarily stored in the image buffer 27.

In the binarization circuits 11a and 11b (11c), the morphological operation image shown in FIG.
FIG. 5 (K) and FIG. 5 are obtained by binarizing (I) and FIG. 5 (J).
The binarized image shown in (L) is obtained. The threshold value for binarization is higher in FIG. 5 (J) than in FIG. 5 (I). FIG.
(K) is a binarized version of FIG. 5 (I) with a relatively high threshold value, in which a defect with a high density and a small area is detected, and a defect with a low density and a large area is not detected. On the other hand, FIG.
(L) is a binarized version of FIG. 5 (J) with a relatively low threshold value. A defect having a low density and a large area is detected, and a defect having a high density and a small area is already removed in the morphological operation described above. And of course not detected.

As described above, the morphology operation circuit 9
By combining a, 9b, and 9c with the binarization circuits 11a, 11b, and 11c, a defective area to be detected can be classified based on both the difference between the normal density value and the size of the defective area. In FIG. 4, three combinations A, B, and C indicate that a defective area to be detected can be set based on the area of the constituent function in the morphological operation called opening and the threshold value in the binarization. That is, in each of the combinations A, B, and C in FIG. 4, the area of the defect detected by dividing the area by the two boundaries is the area away from the origin in the upper right direction of the two boundaries in FIG. It is.

Therefore, the area of the logical sum of the defective areas detected by each of the combination A, the combination B, and the combination C is a detectable defective area. In FIG. 4, the detectable defective area substantially includes a light defect area, a medium defect area, and a heavy defect area, and does not include a normal area. If more combinations than the three combinations A, B, and C shown in FIG. 4 are used, the degree of coincidence between the detectable defective area and each defective area can be made more accurate.

In FIG. 4, the area B and / or the density value of the constitutive function are larger in the combination B than in the combination A and larger in the combination C than in the combination B. In addition, as for the binarization threshold, the combination B is smaller than the combination A and the combination C is smaller than the combination B. That is, the failure area at the intersection of the two boundaries is approximately equal to the area of the configuration function, and the failure level at the intersection of the two boundaries is approximately equal to the threshold for binarization. As shown in FIG. 4, the predetermined threshold value corresponding to each of the constituent functions is set such that a small threshold value corresponds to the constituent function having a large area and / or a density value. The tendency between general recognition and detected defects can be matched.

FIG. 9 is a diagram showing the classification of defective areas performed by the determination circuit 13 when three combinations are used. As shown in FIG. 9, the number of defective areas depends on the difference between the normal density value and the area size of the defective area. 0 to
No. It is divided into seven. However, no. The region of 0 is within the allowable range and is not determined to be defective. The detected defect is N
o. 1 to No. Which of the regions No. 7 belongs can be determined based on which of the combination A, the combination B, and the combination C belongs to which region. Further, when the weight of “1” is assigned to the failure of the combination A, “2” is assigned to the failure of the combination B, and the weight of “4” is assigned to the failure of the combination C, and the addition is performed, 1 to No. 7
Matches the category. Whether the defect is a heavy defect or a light defect can be determined based on the magnitude of the numerical value.
In this way, the determination circuit 13 determines the defect type value indicating the area to which the detected defect belongs, for example, No. 0-No.
7 is output.

The control output circuit 14 inputs the defect type value and outputs a corresponding operation amount. For example, if the detected defect is No. 1, No. 2, No. If the defect is in the area of No. 4, it is determined that the defect is a minor defect and an alarm is output. In addition, the detected defect is No. 3, No. 6, No. If the defect is in the area of No. 7, it is determined that the defect is a medium defect or a heavy defect, and a control output for removing the defective printed matter in the discharge unit of the printing press is performed.

Next, the morphological operation will be described. The morphological operation in the present invention is a multi-valued morphological operation. When a function f defined on E ^N represents the density of an image, the function f is called a multi-valued image of E ^N. For example, a two-dimensional digital image (Z ² )
Is a value of the function f (m, n) when the pixel (m, n) represents a density value. Since the shadow transformation and the surface transformation are used to define the multi-valued morphology operation, first, the shadow transformation and the surface transformation are defined. (Definition 1) Shading conversion: The shading of the function f is described as U [f] and is defined by the following equation 1.

(Equation 1) (Definition 2) Surface transformation: The surface transformation of the closed set A on E ^N is
S [A] and is defined by the following equation (2). As is clear from Equations 1 and 2, S [A]: F → E.

(Equation 2)

As described above, the morphological operations are "dilation", "erosion", "opening" and "closing".
Is composed of four basic operations. First, multi-valued dilation
And multi-valued erosion. In the following description, the function f is referred to as a target image, and the function g is referred to as a constituent function (in the case of binary, also referred to as a constituent element). Like the function f, the configuration function can be regarded as an image. (Definition 3) Multi-valued dilation: The dilation of f by g is
It is defined by the following equation (3).

(Equation 3) (Definition 4) Multi-valued erosion: The erosion of f by g is
It is defined by the following equation (4).

(Equation 4)

Next, the definition of multivalued opening and multivalued closing will be described. (Definition 5) Multi-valued opening: The opening of f by g is
It is defined by the following equation (5).

(Equation 5) (Definition 6) Multi-valued closing: Closing of f by g is
It is defined by the following equation (6).

(Equation 6)

[0040]

As described above, according to the present invention, there is provided an image inspection method and apparatus for judging pass / fail based on an inspection criterion which treats both the difference from a normal density value and the area size of a defect as one. . Further, according to the present invention in which the pass / fail judgment is made based on all the calculation results using a plurality of predetermined configuration functions, the degree of coincidence between the inspection reference and the desired reference is high. Further, according to the present invention, the grayscale image is an image obtained by calculating at least the absolute value of the difference based on the reference image and the inspection target image, and the morphological calculation process is simplified. In the pass / fail judgment step, the type of the defect is determined based on both the binarized image corresponding to each of the constituent functions and the defect type based on both the attribute of the difference between the normal density value of the defect and the size of the area, and the defect type value is determined. According to the present invention, it is possible to determine whether the defect detected by the defect type value is a heavy defect or a light defect. Further, according to the present invention including the control process of outputting the operation amount corresponding to the defect type value, it is possible to automatically perform an appropriate operation such as an alarm or a print stop according to the defect type value by the control process. Further, according to the present invention, in which the morphological operation is an opening operation, it is possible to obtain an operation image in a form suitable for a simple operation. According to the present invention, the predetermined threshold value corresponding to each of the constituent functions has a large area and / or a small value corresponds to a small density value. Tend to be the same as defective.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of an apparatus (image inspection apparatus) for implementing an image inspection method of the present invention.

FIG. 2 is a pictorial diagram showing a reference image (A), a dead zone image (B), an inspection target image (C), and a simple difference absolute value image (D).

FIG. 3 shows the absolute difference value (E) of the substantial difference, 2 at different thresholds
It is a pictorial diagram showing the binarized images (F), (G), and (H) obtained by performing the binarization.

FIG. 4 is a diagram showing a general tendency of human recognition of a defective density (difference from a normal density) and a defective area.

FIG. 5 is a diagram showing morphological operation images (I) and (J) and binarized images (K) and (L).

FIG. 6 is a diagram illustrating an example of a configuration function.

FIG. 7 is an explanatory diagram of a method for performing a morphological operation.

FIG. 8 is a block diagram illustrating an example of an arithmetic device that performs an opening operation that is one of morphological operations.

FIG. 9 is a diagram illustrating a defect area division performed by a determination circuit when three combinations are used.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Image input camera 2 Image input buffer 3 Synchronization circuit 4 Reference image memory 5 Dead zone operation circuit 6 Dead zone image memory 7 Reference image update / dead zone image creation command circuit 8 Difference. Absolute value conversion circuit 9a, 9b, 9c Morphological operation circuit 10 Configuration function setting circuit 11a, 11b, 11c Binarization circuit 12 Binarization threshold value setting circuit 13 Judgment circuit 14 Control output circuit 21, 24, 27 Image buffer 22, 25 Configuration function control unit 23 Minimum value calculation unit 24 Maximum value calculation unit

Claims (8)

[Claims] An image inspection method for comparing and inspecting a reference image and an image to be inspected, comprising the steps of: performing a morphology operation on a grayscale image using a predetermined configuration function to obtain a morphology operation image; A binarization step of performing a binarization operation on the morphology operation image using a threshold to obtain a binarized image; and a pass / fail judgment step of performing pass / fail judgment based on the binarized image. Characteristic image inspection method. 2. The morphological operation step of performing a morphological operation on a grayscale image using a plurality of predetermined constituent functions to obtain a morphological operation image corresponding to each of the constituent functions. The process performs a binarization operation on a morphological operation image corresponding to each of the constituent functions using a predetermined threshold value corresponding to each of the constituent functions, and generates a binary image corresponding to each of the constituent functions. The pass / fail judgment step is performed by a process corresponding to each of the constituent functions.
The image inspection method according to claim 1, wherein a pass / fail judgment is made based on all of the digitized images. 3. The image inspection method according to claim 1, wherein the grayscale image is an image obtained by calculating at least an absolute value of the difference based on the reference image and the image to be inspected. 4. The pass / fail judgment step includes, based on all of the binarized images corresponding to each of the constituent functions, determining a defect type from both defect attributes of a difference between a normal density value and an area size of the defect. The image inspection method according to any one of claims 1 to 3, wherein the determination is performed to obtain a defect type value. 5. The image inspection method according to claim 1, further comprising a control step of outputting an operation amount corresponding to the defect type value. 6. The image inspection method according to claim 1, wherein said morphological operation is an opening operation. 7. A predetermined threshold value corresponding to each of the constituent functions is such that a small threshold value corresponds to a constituent function having a large area and / or a density value.
7. The image inspection method according to any one of claims 6 to 6. 8. An image inspection apparatus for inspecting a reference image by comparing an image to be inspected with a reference image, wherein a morphological operation is performed on a grayscale image using a plurality of predetermined constituent functions, and a morphological operation corresponding to each of the constituent functions is performed. A morphology operation unit for obtaining an image, and performing a binarization operation on the morphology operation image corresponding to each of the constituent functions using a predetermined threshold value corresponding to each of the constituent functions, and corresponding to each of the constituent functions. 2
An image inspection apparatus comprising: a binarizing unit that obtains a binarized image; and a pass / fail determination unit that determines pass / fail based on all of the binarized images corresponding to each of the constituent functions.