Image processing device and image processing method

Abstract

To make it possible to perform alignment even when there are few feature points of a reference image and to improve inspection accuracy.SOLUTION: An image processing apparatus according to the present invention includes: image acquisition means for acquiring a reference image that is a print result to be referenced and a target image that is an inspection target of the print result; detection means for detecting feature points from the reference image; alignment means for performing alignment by obtaining a transformation between the reference image and the target image based on vertices of the reference image and the target image, and the feature points; and inspection means for inspecting the target image based on a comparison between the aligned reference image and the target image. The alignment means calculates a translation component of the transformation based on the feature points, and calculates at least one of a rotation component and a scale component of the transformation based on the vertices.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a technique for inspecting whether or not there are defects in printed matter output by a printing device. [Background technology]

[0002] Inspection work is sometimes carried out to ensure that printed matter is free of defects and of acceptable quality. For example, image data of a good printed matter (hereafter also referred to as a reference image) is prepared in advance. Next, image data of the printed matter to be inspected (hereafter also referred to as a target image) is obtained by scanning or photographing, etc. The inspection is then carried out by comparing these two sets of image data.

[0003] To perform a more accurate comparison inspection, it is necessary to align the printed matter. One method of alignment is to extract edge information from the image, define the points where the edges are concentrated as feature points, and then align the matter based on the detected feature points.

[0004] However, in some images, feature points may not be detected, or even if they are detected, there may be very few. In such cases, even if registration is performed based on feature points, the registration accuracy may be low. Specifically, when registration is performed based on only one feature point, registration by translation is possible, but registration by rotation or scale change is not possible. Furthermore, when registration is performed based on only two feature points, registration by translation, rotation, and scale change is possible, but only scale changes that are equal in both length and width are possible. In a configuration in which paper is fed by a transport device and the paper is scanned by a line scanner to acquire an image, accurate registration cannot be performed because the vertical and horizontal scales are different.

[0005] If a comparison inspection is performed when the alignment accuracy is low, there is a possibility that a non-defective printed matter will be erroneously determined to be defective. Here, Patent Document 1 describes a method of performing an inspection based on the number of feature points without alignment when the number of feature points is small. Specifically, it describes a method of determining that a defect has occurred due to dirt or the like adhering to the printed matter, and thus that the printed matter is defective. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] Japanese Patent Application Laid-Open No. 2013-101015 Summary of the Invention [Problem to be solved by the invention]

[0007] With the method described in Patent Document 1, if the contrast of the defect is low, the defect may not be detected as a feature point. Furthermore, inspection based on the number of feature points makes it impossible to determine the size or position of a detected defect. Furthermore, because the method described in Patent Document 1 does not perform alignment, it is not possible to inspect for print position misalignment, and therefore it is not possible to detect print position misalignment abnormalities.

[0008] The present invention aims to enable alignment even when the reference image has few feature points, thereby improving the accuracy of inspection. [Means for solving the problem]

[0009] The image processing device according to the present invention includes an image acquisition unit that acquires a reference image that is a standard for inspection and a target image obtained by reading a printed matter, a first detection unit that detects feature points from the reference image, and Target image background area and the printed matter Paper used for of Paper By tracking the pixels corresponding to the boundary with the region, Papera second detecting means for detecting the vertex of the Paper and an inspection means for inspecting the target image based on the aligned reference image and target image, wherein the alignment means calculates a translation component of the transformation based on the feature points, and calculates at least one of a rotation component and a scale component of the transformation based on the vertices. The second detection means performs a binarization process on the target image so that pixel values ​​of the background region become a first value and pixel values ​​of the paper region become a second value different from the first value, tracks pixels that are the boundary between the region corresponding to the first value and the region corresponding to the second value, extracts a locus, and detects parts of the outline of the paper that correspond to the locus and form an angle equal to or greater than a predetermined angle as vertices of the paper. It is characterized by: [Effects of the Invention]

[0010] According to the present invention, even if the reference image has a small number of feature points, alignment can be performed, and highly accurate inspection can be performed. [Brief explanation of the drawings]

[0011] [Figure 1] FIG. 1 is a diagram showing the overall configuration of a printing system including an image processing apparatus. [Figure 2] FIG. 1 is a block diagram showing the configuration of an image processing apparatus according to a first embodiment. [Figure 3] 1 is a flowchart of the first embodiment. [Figure 4] 10 is a flowchart of alignment processing according to the first embodiment. [Figure 5] Schematic diagram of the reference and target images. [Figure 6] Overview of the reference and target images (when there are few feature points). [Figure 7] FIG. 10 is a block diagram showing the configuration of an image processing apparatus according to a second embodiment. [Figure 8] 10 is a flowchart of a second embodiment. [Figure 9] FIG. 10 is a diagram illustrating an outline of feature point distribution determination processing according to the second embodiment. [Figure 10] 10 is a flowchart of alignment processing according to the second embodiment. [Figure 11] 10 is a flowchart of a first modified example. [Figure 12] 10 is a flowchart of variant example 5. [Figure 13]10 is a flowchart of variant example 6. DETAILED DESCRIPTION OF THE INVENTION

[0012] The present embodiment will be described below with reference to the drawings. Note that the following embodiment does not necessarily limit the present invention. Furthermore, not all of the combinations of features described in the present embodiment are necessarily essential to the solution of the present invention, and multiple features may be combined arbitrarily. Furthermore, in the accompanying drawings, the same reference numerals are used to designate identical or similar components, and redundant explanations will be omitted. Furthermore, each process (step) in the flowchart is indicated by a symbol beginning with S.

[0013] In this embodiment, the image processing device according to this embodiment inspects the presence or absence of defects in the inspection target medium on which printing has been performed (hereinafter also referred to as the print result). Furthermore, image data representing a printout without defects is referred to as the reference image, and scanned image data of the print result to be inspected is referred to as the target image. [Example]

[0014] In this embodiment, the translation components and rotation / scale components of alignment are calculated from different alignment reference point groups depending on the number of feature points. Here, the point group used for alignment, such as feature points in the image or the vertices of the paper, is called the alignment reference point group. In this embodiment, when the number of feature points is one or two, alignment is performed using both the feature points and the vertices of the paper, and a comparison inspection is performed using the aligned image. This makes it possible to detect low-contrast defects as well.

[0015] (Printing system configuration) FIG. 1 shows an example of the overall configuration of a printing system that outputs and inspects printed materials, including an image processing device 100 to which the present invention is applied. The printing system of this embodiment includes the image processing device 100, a printing server 180, and a printing device 190. The printing server 180 generates a print job for a document to be printed and submits the print job to the printing device 190. The printing device 190 forms an image on a recording medium (paper) based on the print job submitted from the printing server 180. The printing device 190 has a paper feed unit 191, and the user loads printing paper into the paper feed unit in advance. When a print job is submitted, the printing device 190 transports the printing paper loaded in the paper feed unit 191 along a transport path 192, forms an image on one or both sides of the paper, and sends it to the image processing device 100.

[0016] The image processing device 100 of the present invention performs an inspection process to check for defects on printed matter that has been formed by a printing device 190 and sent via a conveying path 192. That is, the image processing device 100 functions as an inspection processing device. The image processing device 100 internally has a CPU 101, RAM 102, ROM 103, a storage device 104, and an image reading device 105. It also has an interface (I / F) 106 with the printing device, a general-purpose interface (I / F) 107, a user interface (UI) panel 108, and a main bus 109. It also has a conveying path 110 for printed matter connected to the conveying path 192 of the printing device 190, an output tray 111 for printed matter that has passed inspection, and an output tray 112 for printed matter that has failed inspection.

[0017] The CPU 101 is a processor that controls all components of the image processing device 100. The RAM 102 functions as the CPU 101's main memory, work area, etc. The ROM 103 stores programs executed by the CPU 101. The main memory device 104 stores applications executed by the CPU 101 and data used in image processing. The image reading device (scanner) 105 can read one or both sides of a printed document sent from the printing device on the conveyance path 110 and acquire the image data. In this embodiment, the image reading device 105 is a line scanner. The printing device I / F 106 is connected to the printing device 190, allowing the two devices to synchronize the processing timing of the printed document with the printing device 190 and to communicate their operating statuses. The general-purpose I / F 107 is a serial bus interface such as USB or IEEE 1394, allowing the user to export data such as logs. The UI panel 108 displays a user interface on an LCD screen, displaying the current status and settings to the user. In addition, print settings such as paper size and type are received from the user via, for example, a mouse or keyboard (not shown) connected to the general-purpose I / F 107. A main bus 109 connects the various components of the image processing device 100. In addition, although not shown in FIG. 2, the CPU 101 can operate various components inside the image processing device 100 and the printing system in response to instructions. For example, it can synchronize the movement of transport paths, or switch whether printed materials are sent to a pass output tray 111 or a fail output tray 112 depending on the inspection results.

[0018] Overall, the image processing device 100 transports printed materials sent from the printing device 190 along a transport path 110, while performing the inspection process described below based on the image data of the printed materials read by the image reading device 105. If the printed materials pass the inspection, they are transported to a pass output tray 111, and if not, they are transported to a fail output tray 112. In this way, only those whose quality has been confirmed can be collected in the output tray 111 for delivery.

[0019] (Configuration of image processing device) 2 shows the configuration of the image processing device 100. The image processing device 100 includes a reference image acquisition unit 201, a feature point acquisition unit 202, a target image acquisition unit 203, a position adjustment unit 204, and an inspection unit 205.

[0020] The reference image acquisition unit 201 acquires a reference image that serves as a reference for inspection. The feature point acquisition unit 202 detects feature points to be used for alignment from the reference image. The target image acquisition unit 203 acquires a target image to be inspected, which is obtained by the image reading device 105 reading a printed material on the conveying path 110. The alignment unit 204 aligns the reference image and the target image. The inspection unit 205 performs a comparative inspection of the printed image using the aligned reference image and target image.

[0021] (Processing performed by the image processing device) The flow of processing in the first embodiment executed by the image processing device 100 will be described with reference to Fig. 3. Fig. 3 is a flowchart showing the flow of processing in the entire device.

[0022] In S1010, the reference image acquisition unit 201 acquires a reference image that serves as a reference for the inspection. The reference image is assumed to be a reference image that has been created in advance by scanning a printed matter that has been visually confirmed to be free of stains and that has been stored in the storage device 104.

[0023] In S1020, the feature point acquisition unit 202 detects feature points from the reference image acquired in S1010. In this embodiment, corners of the print area of ​​the paper are detected as feature points. The so-called Harris corner detection algorithm is used to detect the corners, but the method is not limited to this. Also, although corners are detected as feature points in this embodiment, pixels having other characteristics may also be detected as feature points.

[0024] In S1030 , the target image acquisition unit 203 acquires a target image obtained by the image reading device 105 reading the printed matter on the conveyance path 110 .

[0025] In S1040, the registration unit 204 aligns the reference image and the target image using the feature points acquired in S1020. Details of the registration process will be described later.

[0026] In S1050, the inspection unit 205 calculates the difference between the reference image and the aligned target image. Pixels for which the difference is greater than a predetermined value are detected as defective. If no defect is detected in the target image, the corresponding printed matter is deemed acceptable; if a defect is detected in the target image, the corresponding printed matter is deemed acceptable.

[0027] In S1060, if printing by the printing device 190 is completed, the process ends. If printing is to continue, the process returns to S1030 and continues.

[0028] (Details of the alignment process) The alignment process performed in S1040 will be described with reference to the flowchart in FIG.

[0029] In S1041, the registration unit 204 acquires corresponding points on the target image that correspond to the feature points acquired in S1020. FIG. 5 is a diagram showing the relationship between the reference image and the target image. A reference image 501 has four feature points 502. The registration unit 204 first sets templates 503 of a specified size centered on each of these feature points on the reference image 501. Then, the registration unit 204 performs template matching on the target image 511 using the set templates to acquire corresponding points 512 on the target image 511 that correspond to the feature points 502.

[0030] In S1042, if the number of feature points acquired in S1020 is three or more, the process proceeds to S1043, and if the number is two or less, the process proceeds to S1044.

[0031] In S1043, the registration unit 204 calculates a transformation X from the target image 511 to the reference image 501 based on the feature points 502 on the reference image 501 and the corresponding points 512 on the target image 511. The transformation X in this embodiment is an affine transformation, which represents a transformation that combines translation, rotation, and scale from the target image 511 to the reference image 501. The coordinates of the four feature points 502 on the reference image 501 are (a x 1,a y 1)~(a x 4,a y 4), the coordinates of four corresponding points 512 on the target image 511 (b x 1,b y 1) to (b x 4,b y 4) is written as follows.

[0032] Here, the transformation from corresponding points on the target image to the reference image is expressed as follows. The number of feature points is n. In this embodiment, X is calculated when n=4. The alignment unit 204 calculates the transformation X by multiplying the inverse matrix of B in (Equation 1) from the right of A. If A and B in (Equation 1) are not square matrices, the alignment unit 204 finds the Moore-Penrose pseudo-inverse matrix for matrix B and calculates X. The component x of X 13 ,x 23 The translation component is x 11 ,x 12 ,x 21 ,x 22 , represent the rotation and scale components. A=BX (Formula 1) however,

[0033]

number

[0034] In S1044, the registration unit 204 detects the vertices of the paper in the reference image and the target image. To detect the vertices, the image is first binarized so that the background area of ​​the target image has a pixel value of 0 and the paper area has a pixel value of 1. Next, the pixels that form the boundary between pixel values ​​0 and 1 are tracked and a trajectory is extracted. Then, parts where the contour forms an angle greater than a predetermined angle are detected as vertices. Figure 6 shows examples of a reference image and a target image with two or fewer feature points. Reference image 601 and target image 612 have two feature points 602 and 612. Vertex 603 is detected from reference image 601, and vertex 613 is detected from target image 611.

[0035] In S1045, the registration unit 204 calculates a transformation X' from the target image 611 to the reference image 601 based on the vertices of the paper detected in S1044. The number of vertices of the paper is set to m. In this embodiment, X' is calculated with m=4. The transformation X' is calculated in the same way as in S1043. However, if the coordinates of the vertices 603 on the reference image 601 are (a x '1,a y '1)~(a x '4,a y '4), the coordinates of the corresponding point 613 on the target image 611 are (b x '1,b y '1)~(b x '4,b y It is written as '4). A'=B'X' (Formula 2) however,

[0036]

number

[0037] In S1046, if the number of feature points acquired in S1020 is one or more, the process proceeds to S1047; if the number is zero, X=X' and the process proceeds to S1048.

[0038] In S1047, the registration unit 204 calculates a translation component T from the target image 611 to the reference image 601 based on the feature points 602 on the reference image 601 and the corresponding points 612 on the target image 611. The translation amount T can be calculated by the following formula. Here, the coordinates of the n feature points 602 on the reference image 601 are (a x 1,a y 1)~(a x n,a y n), the coordinates of n corresponding points 612 on the target image 611 are (b x 1,b y 1)~(b x n,b y It is written as n).

[0039]

number

[0040] Next, a transformation X'' is calculated by combining the translation component T calculated based on the feature points and the rotation and scale components calculated based on the vertices of the paper. Here, the transformation X'' is obtained by replacing the translation component of the transformation X' calculated in S1045 with T. x in the transformation X' 13 ',x 23 Since ' represents the translation component, the transformation X'' can be obtained as follows.

[0041]

number

[0042] Then, set X=X'' to complete the calculation of the transformation.

[0043] In S1048, the registration unit 204 transforms the target image using the transformation X determined according to the number of feature points. Through this transformation, the target image that is registered with the reference image is obtained.

[0044] In this way, in this embodiment, when there is one or two feature points, the translation component is calculated based on the feature points, and the rotation and scale components are calculated based on the vertices of the paper. Therefore, when there is one feature point, the rotation and scale components can be calculated, and when there are two feature points, the scale components can be calculated more accurately.

[0045] The above process enables highly accurate alignment even when there are only a few feature points, which allows for comparative inspection to detect low-contrast defects and determine the size and location of detected defects. [Example]

[0046] When feature points are arranged in a straight line, even if there are three or more feature points, the amount of information is only the same as when there are two feature points. Therefore, using only feature points as the alignment reference point group may result in a decrease in alignment accuracy. Therefore, in this embodiment, the translation components and rotation / scale components of alignment are obtained from different alignment reference point groups depending on the distribution of feature points. In this embodiment, even when there are three or more feature points, if the feature points are distributed in a straight line, alignment is performed using the feature points and the vertices of the paper, and a comparison inspection is performed using the aligned image. This makes it possible to improve alignment accuracy when feature points are distributed in a straight line.

[0047] (Configuration of image processing device) 7 shows the configuration of an image processing device 100 of this embodiment. The image processing device 100 has a reference image acquisition unit 701, a feature point acquisition unit 702, a feature point distribution determination unit 703, a target image acquisition unit 704, a positioning unit 705, and an inspection unit 706. The reference image acquisition unit 701, the feature point acquisition unit 702, the target image acquisition unit 704, the positioning unit 705, and the inspection unit 706 are the same as those in the first embodiment, and therefore will not be described here.

[0048] The feature point distribution determining unit 703 determines the distribution of the feature points acquired by the feature point acquiring unit 702. In this embodiment, it is determined whether the feature points are distributed on a straight line.

[0049] (Processing performed by the image processing device) The flow of processing in the second embodiment executed by the image processing device 100 will be described with reference to Fig. 8. Fig. 8 is a flowchart showing the flow of processing in the entire device. The following description will be limited to differences from the first embodiment, and descriptions of the same parts will be omitted.

[0050] In S1130, the feature point distribution determination unit 703 determines whether the feature points acquired by the feature point acquisition unit 702 are distributed on a straight line. Fig. 9 shows an overview of the straight line distribution determination process. A reference image 901 has three feature points 902. First, a regression line 903 of the feature points 902 is calculated by the least squares method. The equation of the line is found as follows: y=sx+t

[0051]

number

[0052] t=μ y -sμ x where μ x ,μ y represents the mean value of x and y of the feature points, σx represents the variance of x of the feature points, and Cov represents the covariance of the feature points.

[0053] Next, the maximum error M between the line and each feature point is calculated as follows. In this embodiment, the error in the y direction is calculated. However, if the coordinates of each feature point are (x i ,y i In the example of FIG. 9, the maximum error value M is 904. M=max(|y i -sx i -t|) (Formula 6) If the calculated M is equal to or smaller than a predetermined threshold, it is determined that the feature points are distributed on a straight line.

[0054] In S1150, the registration unit 705 aligns the reference image and the target image based on the feature points acquired in S1120 and the determination result in S1130. Fig. 10 shows a flowchart of the registration process in the second embodiment.

[0055] In S1153, the alignment unit 705 changes the processing based on the determination result in S1130. If it is determined that the feature points are distributed on a straight line, the process proceeds to S1155, and if not, the process proceeds to S1154.

[0056] In this way, in this embodiment, when feature points are distributed on a straight line, the translation components are calculated based on the feature points, and the rotation and scale components are calculated based on the vertices of the paper. This makes it possible to more accurately determine the scale components.

[0057] The above process enables highly accurate alignment even when feature points are distributed on a straight line, allowing for comparative inspection to detect low-contrast defects and determine the size and location of detected defects.

[0058] [Variation 1] In the first and second embodiments, when the number of feature points or the distribution of feature points satisfies a predetermined condition, the translation components are calculated from the feature points and the rotation and scale components are calculated from the vertices of the paper. However, the combination of the alignment reference point group and the transformation components is not limited to this. For example, the translation and rotation components may be calculated from the feature points and the scale may be calculated from the vertices of the paper.

[0059] A flowchart of the alignment process of this modified example corresponding to the first embodiment is shown in Fig. 11. The following description will be limited to the differences from the first embodiment, and the same parts will not be described.

[0060] In S1246, the alignment unit 204 calculates the scale conversion components from the target image to the reference image. First, the length of each side of the paper is calculated from the coordinates of the vertices of each image. Next, the scale components are calculated from the ratio of the side lengths of the reference image and the target image.

[0061] In S1247, the registration unit 204 converts the feature point on the target image using the scale component calculated in S1246. The conversion from the feature point B before conversion to the feature point B' after conversion using the scale component is performed using the following formula.x is the scale component in the x direction, s y Let be the scale component in the y direction. B'=BX' (Formula 7)

[0062]

number

[0063] In S1248, a transformation X'' from the feature point B' transformed in S1247 to the feature point A of the reference image is calculated. A=B'X''...(Formula 8)

[0064]

number

[0065] Then, since the calculated X'' represents the translation and rotation components, it is combined with the scale component calculated in S1247 to calculate the final transformation X.

[0066]

number

[0067] [Variation 2] In the first embodiment, the translation components are calculated from the feature points and the rotation and scale components are calculated from the vertices of the paper only when there are one or two feature points, but the condition based on the number of feature points is not limited to this. When the number of feature points is 1 to n, the translation components may be calculated from the feature points and the rotation and scale components may be calculated from the vertices of the paper. Alternatively, regardless of the conditions, the translation components may be calculated from the feature points and the rotation and scale components may be calculated from the vertices of the paper.

[0068] [Variation 3] In the first and second embodiments, the transformation X is calculated as an affine transformation, but the transformation is not limited to this. For example, a projective transformation may be calculated as the transformation X.

[0069] [Variation 4] In the first and second embodiments, feature points are detected and acquired from the reference image, but feature points previously set by the user may also be used. For example, feature points may be set by the user via the UI panel 108 based on the reference image and used.

[0070] [Variation 5] For the transformation components calculated based on the feature points in the first and second embodiments, if certain conditions are met, translation components may be calculated based on the feature points, and rotation and scale components may be calculated based on the vertices of the paper. A flowchart of the alignment process for this modified embodiment is shown in Figure 12. The following description will be limited to the differences from the first embodiment, and a description of the same parts will be omitted.

[0071] In S1354, it is determined whether the transformation components found based on the feature points satisfy predetermined conditions. In this modification, it is determined whether the horizontal and vertical scale components of the transformation are equal to or greater than predetermined values. If the determination result shows that the conditions are satisfied, the process proceeds to S1355; if not, the process proceeds to S1359. Note that the predetermined conditions are not limited to this. For example, it may be determined whether the rotation component is equal to or greater than a predetermined angle.

[0072] [Variation 6] In addition to the comparison inspection, the inspection unit 205 may also inspect for print position misalignment of a pattern (pattern misalignment inspection). A flowchart of the inspection process when inspecting for pattern misalignment is shown in Figure 13. The following explanation will be limited to the differences from the first embodiment, and explanations of the same parts will be omitted.

[0073] In S1051, if there is one or more feature points, the process proceeds to S1052, and if there is zero feature points, the process proceeds to S1054.

[0074] In S1052, the inspection unit 205 inspects the print position misalignment of the image. First, if the transformation based on the vertices of the paper was not calculated during the alignment process, the vertices of the paper are detected and the transformation based on the vertices of the paper is calculated. Next, the difference between the translation component calculated based on the vertices of the paper and the translation component calculated based on the feature points is calculated. If the calculated difference is greater than a predetermined threshold, it is determined that there is an abnormal print position misalignment.

[0075] The print position deviation inspection process is not limited to this method. For example, the coordinates of the vertices of the target image may be transformed using a vertex-based transformation and a feature-point-based transformation, and if the difference between the two is greater than a predetermined threshold, it may be determined that there is a print position deviation abnormality.

[0076] In S1053, if it is determined in S1052 that there is no abnormality in the print position deviation, the process proceeds to S1054. If it is determined that there is an abnormality in the print position deviation, the inspection process ends.

[0077] By combining the first embodiment with this modified example, it is possible to perform highly accurate alignment and detect abnormal print position deviation even when there are few feature points.

[0078] The present invention is not limited to the above-described embodiments, and may be implemented by combining the elements and concepts described in each embodiment.

[0079] The present invention can also be realized by executing the following process: software (programs) that realize the functions of the above-described embodiments are supplied to a system or device via a network or various storage media, and the computer (or CPU, MPU, etc.) of the system or device reads and executes the programs. [Explanation of symbols]

[0080] 201 Reference image acquisition unit 202 Feature point acquisition unit 203 Target image acquisition unit 204 Alignment section 205 Inspection Department

Claims

1. an image acquisition means for acquiring a reference image that is a standard for inspection and a target image obtained by reading a printed matter; a first detection means for detecting feature points from the reference image; a second detection means for detecting the vertices of the paper by tracking pixels in the target image that correspond to the boundary between a background area of ​​the target image and a paper area of ​​the paper used for the printed matter; a registration means for performing registration between the reference image and the target image by calculating a transformation for performing registration between the reference image and the target image based on the vertices of the paper and the feature points; an inspection means for inspecting the target image based on the aligned reference image and target image; and the position adjustment means calculates a translation component of the transformation based on the feature points, and calculates at least one of a rotation component and a scale component of the transformation based on the vertices; The second detection means performs a binarization process on the target image so that pixel values ​​of the background region become a first value and pixel values ​​of the paper region become a second value different from the first value, tracks pixels that are the boundary between the region corresponding to the first value and the region corresponding to the second value, extracts a locus, and detects parts of the outline of the paper that correspond to the locus and form an angle equal to or greater than a predetermined angle as vertices of the paper.

1. An image processing device comprising:

2. The positioning means calculates a translation component of the transformation based on the feature points, and calculates a rotation component of the transformation and the scale component based on the vertices.

2. The image processing device according to claim 1, wherein:

3. The positioning means calculates a translation component and a rotation component of the transformation based on the feature points, and calculates a scale component of the transformation based on the vertices.

2. The image processing device according to claim 1, wherein:

4. When the number of the feature points is equal to or less than a predetermined value, the positioning means calculates a translation component of the transformation based on the feature points, and calculates at least one of a rotation component and a scale component of the transformation based on the vertices.

4. The image processing device according to claim 2 or 3.

5. The method further includes a determination means for determining a distribution of the feature points, When the determining means determines that the feature points are distributed on a straight line, the positioning means calculates at least one of a rotation component and a scale component of the transformation based on the vertices.

4. The image processing device according to claim 2 or 3.

6. When the transformation calculated from only the feature points satisfies a predetermined condition, the positioning means calculates a translation component of the transformation based on the feature points, and calculates at least one of a rotation component and a scale component of the transformation based on the vertices.

6. The image processing device according to claim 1, wherein the first and second inputs are input to the image processing device.

7. Further, a pattern position deviation inspection means is provided for inspecting a pattern position deviation between the reference image and the target image, The pattern position deviation inspection means inspects the pattern position deviation when the first detection means detects one or more feature points.

7. The image processing device according to claim 1, wherein the first and second inputs are input to the image processing device.

8. A program for causing a computer to function as each of the means of the image processing apparatus according to any one of claims 1 to 7.

9. an image acquisition step of acquiring a reference image that is a standard for inspection and an object image obtained by reading a printed matter; a first detection step of detecting feature points from the reference image; a second detection step of detecting vertices of the paper by tracking pixels in the target image that correspond to a boundary between a background region of the target image and a paper region of the paper used for the printed matter; a registration step of calculating a transformation for performing registration between the reference image and the target image based on the vertices of the paper and the feature points; an inspection step of inspecting the target image based on the aligned reference image and target image; and In the alignment step, a translation component of the transformation is calculated based on the feature points, and at least one of a rotation component and a scale component of the transformation is calculated based on the vertices; In the second detection step, a binarization process is performed on the target image so that pixel values ​​of the background region become a first value and pixel values ​​of the paper region become a second value different from the first value, pixels that form a boundary between the region corresponding to the first value and the region corresponding to the second value are tracked to extract a locus, and portions of the outline of the paper that correspond to the locus and form an angle equal to or greater than a predetermined angle are detected as vertices of the paper. An image processing method comprising:

Images