JPH0915168A - Inspecting equipment of printed matter - Google Patents

Abstract

PROBLEM TO BE SOLVED: To detect even a minute defect such as a break of a character or a stain. SOLUTION: A reference image picked up from the same patterns printed repeatedly and successively is stored in a reference image memory. An image for inspection is taken in and a positional slippage is detected by a pattern matching part 15. The image for inspection stored in an inspection image memory 13 is put in an address for the one of which the positional slippage is modified, and inputted to an inspecting unit 17 together with the reference image. The inspecting unit 17 prepares difference images of the two images, applies a maximum value filter to either of them, calculates a difference between this and the difference image to which the maximum value filter is not applied and detects a defect therefrom.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a printed matter inspection apparatus for detecting defects such as stains and missing characters on a printed matter in which the same pattern is continuously and repeatedly printed.

[0002]

2. Description of the Related Art As a method for detecting stains or missing characters on a printed material in which the same pattern is repeatedly printed by a plate cylinder, etc., a reference image and an inspection image are compared, and if the difference is equal to or more than a specified value, it is determined as a defect. This is the method of judgment. In comparing the reference image and the inspection image, if there is a positional shift between the reference image and the inspection image, correct comparison cannot be performed. Therefore, some methods for eliminating this shift have been proposed (JP-A-4-33).
9653). As one of the methods, a region in which the density of the inspection image changes abruptly is set as a mask region and is not an inspection target. However, in such a case, most of the label or the like on which characters are printed act as a mask, and it is not possible to detect defects in regions where the characters are missing or the density changes. This mask area is provided to prevent erroneous detection due to the positional deviation between the inspection image and the reference image. That is, when there is a difference between the two images, this difference is regarded as a defect, but when the positions of the images are displaced from each other, the difference due to this displacement is also determined as a defect. Therefore, the outline portion such as a character has a large density change, and if a deviation occurs, it appears as a difference, so that this portion is masked and excluded from the inspection target. This makes it possible to detect when there is dirt, etc., at a position away from the character, but it does not provide detection information for characters such as a part of the character being cut off or a part of the character being dirty. Absent.

[0003]

As described above, in the method using the mask, most of the labels or the like on which characters are printed are masks, and it has been impossible to detect defects in characters or in areas where the density changes. Japanese Patent Publication No. 6-17877 discloses a technique of setting a different density threshold value for each part of the pattern in order to cope with the lateral vibration and expansion / contraction of the inspection image, but the density change of the pattern becomes large. In the vicinity of the part, the allowable value is too large and the defect inspection omission occurs. It is also necessary to set different thresholds for each part.

Japanese Patent Publication No. 7-31134 discloses that a reference image and an inspection image are binarized for matching.
A technique is disclosed in which one image is expanded or contracted and subtracted to detect a defect. However, since the image is a binarized image, even a minute defect cannot be detected by obtaining a difference in dark level for a dark density and a difference in light level for a light density. In addition, when making a binary image, it is necessary to correct the shading such as illumination, external light, and brightness difference in a partial area.

The present invention has been made in view of the above problems, and detects a positional deviation between a reference image and an inspection image, corrects the positional deviation, and expands the printing paper or a quantization error due to digitization during image processing. It is an object of the present invention to provide a printed matter inspection apparatus capable of detecting even minute defects such as missing characters and stains by removing the above.

[0006]

In order to achieve the above object, according to the invention of claim 1, in a printed matter inspection apparatus for inspecting the same pattern repeatedly and repeatedly printed, a storage means for storing a reference image and an inspection object. Correcting means for detecting the positional deviation from the reference image by capturing the inspection image and outputting a corrected image in which this positional deviation is corrected, and applying a maximum value filter to either the reference image or the corrected image, and applying the maximum value filter And a detection unit that binarizes a grayscale image obtained by subtracting the image to which the maximum value filter is applied from the non-applied image and detects the image as a defect image.

With this configuration, the displacement of the inspection image with respect to the reference image is detected, and a corrected image in which the displacement of the inspection image is corrected is obtained. Since the misregistration is detected by digitizing it on a pixel-by-pixel basis, the quantization error associated with the digitization is included in the corrected image in which the misregistration is corrected. In addition, the printing paper is conveyed under tension. Since the position of the reference image and the position of the inspection image are different, if the positions are apart, the degree of expansion and contraction of the printing paper is different. Such a difference between the two images due to the expansion and contraction of the images cannot be eliminated by the alignment, and therefore appears as a shift between the two images. In order to eliminate the deviation between the reference image and the inspection image due to the quantization error and the expansion / contraction of the printing paper, the density change position of one image is slightly widened so that the above two deviations are included in the widened range. By comparing the images, it is possible to eliminate the influence of this shift, that is, to detect this shift as a defect.

Therefore, the maximum value filter is applied to either of the screens for the reference image and the corrected image. In the case of a filter size such as 3 × 3, 5 × 5, 7 × 7, 9 × 9, for example, the maximum value filter sets the density of the pixel at the center to the maximum density of the pixels in the filter. For example, when black characters are printed on a white background with a filter, this maximum value filter causes the black pixels around the characters to become white pixels, and the characters can be erased. On the contrary, when white characters are printed on a black background, the white characters are thick. The maximum value filter is applied to either one of the difference image of the reference image and the corrected image, and the difference from the image not subjected to the maximum value filter is taken, thereby eliminating the defect excluding the influence of the above-described two image shifts. Can be detected.
When the image subjected to the maximum value filter is subtracted from the image not subjected to the maximum value filter, the density difference is usually negative. Negative is zero, and a defect having a difference larger than a predetermined positive value is set as a defect.

According to a second aspect of the present invention, in a printed matter inspection apparatus for inspecting the same pattern repeatedly and repeatedly printed, a positional deviation between the storing means for storing the reference image and the inspection image to be inspected is detected and the reference image is detected. Then, a correction unit that outputs a corrected image in which the positional deviation is corrected, and a difference unit that generates a difference image in which the density difference of two pixels is the density of the central pixel of the two pixels for each of the reference image and the corrected image. And a maximum value filter is applied to either the difference image of the reference image or the difference image of the corrected image, and the grayscale image obtained by subtracting the image subjected to the maximum value filter from the image not subjected to the maximum value filter is set by the threshold value. And a detection unit for binarizing and detecting as a defect image.

According to the second aspect of the present invention, the difference image is generated by the difference means with respect to the reference image and the corrected image with respect to the density difference of two pixels as the density of the central pixel of the two pixels. . The difference image enhances the detection accuracy of the defect image because the position where the density change occurs is emphasized by removing illumination, external light, and gentle change in brightness.

According to the third aspect of the present invention, of the two consecutive images, the first image is the reference image, the second image is the inspection image, and the second image is the reference image. The image is used as the inspection image, and the reference image and the inspection image are similarly defined.

According to the first or second aspect of the present invention, when printing on a material that easily expands and contracts, the pattern expands and contracts due to tension, and therefore the expansion and contraction error with respect to the fixed reference image is large. Therefore, in the invention of claim 3, the first image of two continuous images is used as a reference image, and the next image is used as an inspection image. Since two continuous images have almost the same tension, it can be considered that the expansion and contraction of the printing paper is the same.

According to the invention of claim 4, the detecting means of claim 2 or 3 subtracts the image obtained by applying the maximum value filter to the difference image of the reference image from the difference image of the corrected image to obtain a positive image as a dirty image. To detect as.

An unclean image can be defined as an image that does not exist in the reference image and appears in the inspection image. Since the corrected image has an extra image that is not in the reference image, when the above subtraction is performed, only the extra image that excludes the quantization error and the shift due to expansion and contraction of the printing paper is detected. it can.

In the invention of claim 5, the detecting means of claim 2 or 3 subtracts the image obtained by applying the maximum value filter to the difference image of the corrected image from the difference image of the reference image, and the positive image is missing characters. Detect as an image.

The character missing image can be defined as an image which is present in the reference image but not in the corrected image. For this reason, when the above-mentioned subtraction is performed, it is possible to detect a portion existing only in the reference image, and it is determined that this is a character loss. "Characters" with missing characters are not always letters,
It may be a picture.

In the invention of claim 6, the difference means of claims 2 to 5 generates difference images in the horizontal and vertical directions of the reference image and the corrected image.

When obtaining the difference image, the difference image in the horizontal direction and the difference image in the vertical direction are obtained, and the inventions of claims 2 to 5 are applied to each. As a result, it is possible to detect stains and missing characters in the horizontal and vertical directions.

According to a seventh aspect of the invention, in the second or third aspect of the invention, the difference means generates a difference image in the horizontal and vertical directions of the reference image and the corrected image, and the detection means generates the corrected image. The horizontal difference image that is positive by subtracting the image obtained by applying the maximum value filter to the horizontal difference image of the reference image from the horizontal difference image, and the maximum value filter from the vertical difference image of the corrected image to the vertical difference image of the reference image Vertical smeared image that is positive by subtracting the image multiplied by, and horizontal character missing image that is positive by subtracting the image obtained by applying the maximum value filter to the horizontal difference image of the corrected image from the horizontal difference image of the reference image, Subtract the image obtained by applying the maximum value filter to the vertical difference image of the corrected image from the vertical difference image of the reference image to create a positive vertical character missing image, and To generate an image composed of the logical sum of the image and the vertical stain image and a horizontal character line image and vertical character line image.

With the above configuration, a difference image in the horizontal direction and a difference image in the vertical direction are created for the corrected image and the inspection image, and the horizontal smeared image, the vertical smeared image, the horizontal character missing image,
Since a vertical character missing image is generated and an image composed of the logical sum of these four images is generated, stains and character missing can be combined and displayed on one screen and the position can be detected.

[0021]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is an overall configuration diagram of the embodiment. The same pattern is already repeatedly printed on the printed matter 1 by a plate cylinder or the like (not shown). The roller 2 conveys the printed matter 1 at high speed in the direction indicated by the arrow. One of the rollers 2 is provided with an encoder 3 directly connected to its shaft or via a gear, detects a rotation angle, and outputs it as a pulse. A line sensor camera 4 is installed above the printed matter 1 and captures a one-dimensional image of the printed matter 1. The illumination device 5 illuminates the image pickup position of the printed matter 1. The position sensor 6 detects a register mark or the like printed on the side of each screen of the printed matter 1 to detect the position of a passing pattern. Printing minute defect detection device 7
Detects the defects of the printed matter 1 to the minute, based on the data from the encoder 3, the line sensor camera 4, and the position sensor 6. The CRT 8 is a display device that displays the detection result and the like, and the host computer 9 controls the defect detection processing.

FIG. 2 is a block diagram of the printing micro defect inspection apparatus and its peripheral devices. The camera controller 11 controls the scanning timing of the line sensor camera 4. An image passing through the imaging position of the line sensor camera 4 can be detected from the output of the encoder 3 and the output of the position sensor 6, so the scanning timing is determined according to this image and horizontal scanning is performed. The shading correction unit 12 corrects illumination unevenness of the input image, and performs this correction after converting the input analog data into digital data. The inspection image memory 13 stores the shading-corrected inspection image. Since the line sensor camera 4 obtains a grayscale image, a monochrome camera may be used, but in the case of a color camera, red,
The circuit configurations are the same for the three colors of green and blue. In addition, printing defects for each color can be detected by colorization.
The reference image memory 14 stores the reference image. Inspection image 1
3 and the reference image 14 are provided on each of two surfaces, one for capturing an image, and the other for reading from the inspection unit 17 described later, which are used alternately.

The pattern matching unit 15 registers a partial region of the reference image, for example, a region of 16 × 16 or 32 × 32 pixels as a template, performs alignment with the corresponding region of the inspection image, and Detect the deviation. The address generator 16 determines the address of the data read out from the inspection image memory 13 and the reference image memory 14 to the inspection unit 17, which will be described later, and the inspection corresponding to the address of the reference image. An address whose positional deviation has been corrected for the image is generated, and the positional deviation is corrected.

The inspection unit 17 inputs the reference image and the corrected image in which the displacement of the inspection image has been corrected, and detects a defect by difference processing, maximum value filtering processing, subtraction processing, or the like. The defect image memory 18 stores the defective image detected by the inspection unit 17. Display memory 19 is CRT8
A memory for storing images to be displayed on the inspection image memory 1
3. Read the image data stored in the reference image memory 14 and the defective image memory 18 and store it for display. The display controller 20 selects data to be stored in the display memory 19 and controls the display on the CRT 8.

Next, the operation of the first embodiment will be described. In the present embodiment, a reference image is stored in advance, and the reference image is compared with the reference image to detect stains, missing characters, etc. on the inspection image.
FIG. 3 is an operation flow chart of this embodiment. First, the reference image is stored in the reference image memory 14 (S1). Next, an image representing a part of the reference image is registered as a template in the memory of the pattern matching unit 15 (S2). The template size is, for example, 16 × 16 or 32 × 3.
It is a small image of about 2 pixels. Next, the pattern matching unit 15 takes in the inspection image, detects the position representing the same image as the template, and detects the deviation between the position of the template in the reference image and the position of the same image in the inspection image as the template ( S3).

FIG. 4 is a diagram for explaining a method of detecting the positional deviation. (A) shows the template provided in the reference image and its position. The character G represented by 16 × 16 or 32 × 32 pixels is used as a template, and its position is (X0, Y
It is represented by 0). The area indicated by the broken line is assumed to be a range in which the characters of this template move in the inspection image.
(B) shows the letter G corresponding to the template in the inspection image. The inside of the area indicated by the broken line is checked, the character G is detected, and its position (X, Y) is obtained. Since there is almost no rotational deformation of the inspection image, the positional deviation of the inspection image is represented by the difference between (X0, Y0) and (X, Y). The method of checking the position of the letter G in the inspection image is to move the image of the template in the entire region shown by the broken line, and at each position, the size of the template is 16 × 1.
For each of 6 or 32 × 32 pixels, the density difference from each corresponding pixel of the template is obtained, and the absolute value of the density difference is added for all the pixels of the template. The position where the added value is the minimum is the position of the template image (character G) in the inspection image, and this position is represented by (X, Y). That is, when the density of the pixel at the position i of the template is Si and the density of the pixel corresponding to the position i of the template of the inspection image is Ti, Σ | Si−Ti |
The position at which is minimum is the position of the character G in the inspection image.

When the displacement of the inspection image is detected in this way, when the inspection image is read from the inspection image memory 13 and sent to the inspection unit 17, the address generator 16 generates an address in which the displacement is corrected. , The inspection image is sent to the inspection unit 17 at this address (S
4). This image is called a corrected image. The reference image is sent to the inspection unit 17 at the same time as the corrected image (S
4). The inspection unit 17 creates horizontal and vertical difference images of the reference image and the corrected image (S5). Here, the difference image is a horizontal difference image, and the difference between the densities of the nth pixel and the n + kth pixel of the pixels arranged in the horizontal direction is the density of the central pixel (n + k / 2nd pixel), If there is a density difference between the two pixels, it will be a large value. Making an image of density difference means making a differential image.
The generation of the difference image (S5) is not necessary because it is for emphasizing the density change, but it is usually used. Further, a differential image of the differential image may be obtained, that is, an image in which the features are emphasized by quadratic differentiation may be used.

Next, the maximum value filter is applied to the difference image of the reference image and the difference image of the inspection image (S6). The maximum value filter is a filter that finds the maximum density among the pixels in the filter centering on the target pixel and converts the target pixel to this maximum density. FIG. 5 shows an example when the maximum value filter is applied. As shown in (A), for example, the densities of pixels in a 3 × 3 range centered on the P pixel of the inspection image or the reference image are checked, and as shown in (B), the density of the target pixel is converted to the maximum density. . As a result, for example, when a character is printed in black on a white background, the range of white with high density is large, and the character is thin. On the contrary, when white characters are displayed in black, the characters are thick.

Next, the difference image of the reference image is subtracted from the difference image of the inspection image by the maximum value filter to obtain the difference (S7). This is performed for each of the horizontal difference image and the vertical difference image. When this difference is equal to or larger than the set value, it is detected as a stain defect and a binarized image is generated (S8). In parallel with the processing of the inspection image, the image obtained by applying the maximum value filter to the difference image of the inspection image is subtracted from the difference image of the reference image,
The difference is obtained (S9). The difference image is the horizontal difference image,
And the vertical difference image. When this difference is equal to or larger than the set value, it is detected as a character missing defect and a binarized image is generated (S10).

FIG. 6 shows a procedure for obtaining a stain defect and a character defect defect from the difference between the reference image and the inspection image in the horizontal direction (H direction). The difference between the reference image and the inspection image in the horizontal direction is calculated, and the maximum value filter is applied to each. The value obtained by applying the maximum value filter to the horizontal difference image of the reference image is subtracted from the horizontal difference image of the inspection image. If this value is greater than or equal to the set value, it is determined as a stain defect. Further, the value obtained by applying the maximum value filter to the horizontal difference image of the inspection image is subtracted from the horizontal difference image of the reference image. If this value is greater than or equal to the set value, it is determined as a character missing defect. When a binarized image is created using this predetermined value as a threshold value, stains and character defect defects are clearly represented. The subtraction value on the left side of FIG. 6 indicates a stain defect, and the right side indicates a character missing defect.

FIG. 7 shows a procedure for obtaining a stain defect and a character defect defect from the difference between the reference image and the inspection image in the vertical direction (V direction). The vertical difference between the reference image and the inspection image is calculated, and the maximum value filter is applied to each of them. A value obtained by applying a maximum value filter to the vertical difference image of the reference image is subtracted from the vertical difference image of the inspection image. If this value is greater than or equal to the set value, it is determined as a stain defect. Further, the value obtained by applying the maximum value filter to the vertical difference image of the inspection image is subtracted from the vertical difference image of the reference image. If this value is greater than or equal to the set value, it is determined as a character missing defect. When a binarized image is created using this predetermined value as a threshold value, stains and character defect defects are clearly represented. The subtraction value on the left side of FIG. 7 indicates a stain defect, and the right side indicates a character missing defect.

FIG. 8 is a view showing the contents of each process of FIGS. 6 and 7. The processing content represents either the horizontal direction or the vertical direction. a shows a reference image and b shows an inspection image. "x" shown in the image of b represents what is not in the reference image, that is, stains attached to the inspection image, and "y" represents that which is not in the inspection image, that is, character loss. c represents the difference image of the reference image, and d represents the difference image of the inspection image. e indicates an image obtained by applying the maximum value filter to the difference image of the reference image of c, and the broken line indicates the image of c for reference. f is a value obtained by applying the maximum value filter to the difference image of the inspection image of d, and the broken line is d.
Is shown for reference. g is a value obtained by subtracting the image obtained by applying the maximum value filter to the difference image of the reference image of e from the difference image of the inspection image of d, and extracting only the positive value.
Is shown. h is a value obtained by subtracting the maximum value filter from the difference image of the reference image of c to the difference image of the inspection image of f and extracting only the positive value, and represents the character missing “y”.

In FIG. 8, e and f subjected to the maximum value filter are thicker than the original c and d, and even if the inspection image b deviates from the reference image a within the range of this thickening, subtraction is performed. , The common part of a and b is removed. Further, by making the subtraction result only a positive value (or a predetermined positive threshold value or more), it is possible to detect dirt and character missing separately.

Returning to FIG. 3, the binarized image representing the stain defect and the binarized image representing the character missing defect in the horizontal difference image obtained in FIG. 6 and the stain defect in the vertical difference image obtained in FIG. A composite image is created by the logical sum of the four images of the binarized image represented and the binarized image representing the character missing defect (S11). As a result, dirt and missing characters can be displayed on one screen. A defect position such as dirt or a missing character is detected from this composite image (S1).
2).

FIG. 9 is a diagram showing the procedure for creating a composite image and detecting the defect position. The composite image is created by taking the logical sum of the four binarized images representing the defects obtained in FIGS. 6 and 7. (S2
1). Then, the stains and the character defects shown in this composite image are labeled (S22), and the position of each labeled defect is detected (S23). If such stains or missing characters occur, an error is displayed on the screen and an alarm is issued (S24).

By the procedure described with reference to FIGS. 3 and 9, it is possible to detect minute defects such as dirt and missing characters. Further, it is possible to detect a defect near the point where the concentration changes abruptly. If the maximum value filter is too small, the quantization error and the error due to the elongation of the printing paper cannot be absorbed, but if it is too large, it becomes impossible to detect the defects near the change point where the density suddenly changes.
Usually, a filter of about 3 × 3 to 9 × 9 is used.

Next, a second embodiment will be described. In the case of printed matter printed on a soft material, the tension applied to the printed matter varies depending on the winding position, and there are many expansion and contraction of the material,
The length of the inspection image in the traveling direction may not match the fixed reference image. In such a case, the expansion and contraction of the material can be ignored in the two consecutive images, and the image inside the two consecutive images is used as the reference image and the next image is used as the inspection image. Next, this inspection image becomes the reference image, and the next image becomes the inspection image. In this case, if the image continues as "good, bad, good", the third "good" image is different from the second reference image, and thus "bad", but this is "good".
If it is determined that “good, bad, bad, good” continues, the third “bad” is “good” if it is the same defect as the second reference image, but this is judged as “bad”, The fourth “good” is “defective”, but a criterion is provided to judge this as “good”. The second embodiment can also be used for the inspection of printed matter in which the expansion and contraction of the material is normal.

[0038]

As is apparent from the above description, according to the present invention, after the reference image and the inspection image are aligned with each other, a difference image between the two images is generated and one maximum value filter is applied to the difference image. After that, since the defect is detected by taking the difference from the other difference image, it is possible to detect even minute defects such as stains and missing characters. In addition, by using the preceding image of two consecutive images as the reference image, the subsequent image as the inspection image, and the inspection image as the next reference image for alignment, minute defects such as stains and missing characters can be generated almost in the same manner. Can be detected.

[Brief description of the drawings]

FIG. 1 is an overall configuration diagram of an embodiment.

FIG. 2 is a block diagram of a printing micro defect inspection apparatus and its peripheral devices.

FIG. 3 is an operation flow chart of the first embodiment.

FIG. 4 is a diagram illustrating alignment.

FIG. 5 is a diagram illustrating a maximum value filter.

FIG. 6 is a diagram for explaining detection of horizontal stains and character missing defects.

FIG. 7 is a diagram for explaining detection of vertical stains and character missing defects.

FIG. 8 is a diagram showing a specific example of defect detection.

FIG. 9 is a procedure diagram for creating a composite image of defects and detecting defect positions.

[Explanation of symbols]

 1 Printed Material 2 Roller 3 Encoder 4 Line Sensor Camera 5 Lighting Device 6 Position Sensor 7 Printing Micro Defect Inspection Device 8 CRT 9 Host Computer 13 Inspection Image Memory 14 Reference Image Memory 15 Pattern Matching Section 16 Address Generator 17 Inspection Unit

Claims (7)

[Claims] 1. In a printed matter inspection apparatus for inspecting the same pattern repeatedly and repeatedly printed, a positional deviation between a storage unit for storing a reference image and an inspection image to be inspected is detected and the reference image is detected. Correcting means for outputting a corrected image that has been corrected, a maximum value filter is applied to either the reference image or the corrected image, and a grayscale image obtained by subtracting the image subjected to the maximum value filter from the image not subjected to the maximum value filter is displayed. A printed matter inspection apparatus comprising: a detection unit that binarizes the image with a threshold value and detects the image as a defect image. 2. A printed matter inspection apparatus for inspecting the same pattern repeatedly and repeatedly printed, detecting a positional deviation between a storage means for storing a reference image and an inspection image to be inspected and detecting the reference image. Correction means for outputting a corrected image that has been corrected, difference means for generating a difference image in which the density difference of two pixels is the density of the center pixel of the two pixels for each of the reference image and the corrected image, and the reference image Of the difference image of No. 1 or the difference image of the corrected image, and the grayscale image obtained by subtracting the image on which the maximum value filter is subtracted from the image on which the maximum value filter is not applied is binarized by a threshold value to detect a defect. A printed matter inspection apparatus comprising: a detection unit that detects an image. 3. The first image of the two consecutive images is the reference image, the latter image is the inspection image, the second image is the reference image, and the image next to the latter image is the inspection image. 3. The printed matter inspection apparatus according to claim 1, wherein the reference image and the inspection image are similarly defined. 4. The detection unit subtracts an image obtained by applying a maximum value filter to a difference image of a reference image from a difference image of a corrected image, and detects a positive image as a dirty image. Or the printed matter inspection apparatus described in 3. 5. The detection unit subtracts an image obtained by applying a maximum value filter to a difference image of a corrected image from a difference image of a reference image to detect a positive image as a character missing image. The printed matter inspection apparatus according to 2 or 3. 6. The printed matter inspection apparatus according to claim 2, wherein the difference unit generates a difference image in the horizontal direction and the vertical direction of the reference image and the corrected image. 7. The difference means generates a difference image in the horizontal direction and the vertical direction of the reference image and the corrected image, and the detection means maximizes the horizontal difference image of the reference image from the horizontal difference image of the corrected image. A horizontal smeared image that is positive by subtracting the value-filtered image and a vertical smeared image that is positive by subtracting the maximum-value-filtered image from the vertical difference image of the reference image from the corrected vertical difference image. , A horizontal character missing image that is positive by subtracting the image obtained by applying the maximum value filter to the horizontal difference image of the corrected image from the horizontal difference image of the reference image, and the vertical difference of the corrected image from the vertical difference image of the reference image The image with the maximum value filter applied to the image is subtracted to create a vertical character missing image that is positive, and a horizontal smeared image, a vertical smeared image, a horizontal character missing image, and a vertical sentence are created. The printed matter inspection apparatus according to claim 2 or 3, wherein an image formed by a logical sum of the character missing images is generated.