JP6692052B2 - Image loss detection device and image loss detection method - Google Patents

Description

  Embodiments of the present invention relate to an image loss detection device that detects a loss in an image such as a character or a pattern, and an image loss detection method thereof.

Conventionally, for inspection of the quality of printed characters, for example, a reference character that is a non-defective product is used as a determination reference to compare the print character with the reference character, and whether the print character is good or defective is determined by whether they match. There is. For example, Patent Document 1 discloses a technique for making a determination by giving an allowable width to the line width of a reference character used for comparison.
In this technique, an image pattern that is reduced (or narrowed) inward by a predetermined number of pixels from the line width of a reference character is used as a first reference image and compared with a character image that includes a print character, Detects partial omissions such as "chipped", "perforated", and "cut". Further, an image pattern enlarged (thickened) outward by a predetermined number of pixels from the reference image is used as a second reference image and compared with the character image to detect stains such as “protrusion” of the character image.

Patent Publication No. 4128181

  In the determination according to the above-mentioned Patent Document 1 in which the line width of the reference character is allowed to be compared, when there is a minute point distant from the line of the character or the like, when a pattern in which the reference character is thin is created , That point disappears or shrinks. For this reason, in the photographed inspection image, even if there are no points, they are within the range of good judgment, and there arises a problem that the missing points cannot be detected or the number of missing pixels is detected small.

  In addition, the image pattern enlarged outward by a predetermined number of pixels from the reference image also has a function of a pattern deviation allowable range. From the human appearance, even if a character is printed with a slight misalignment, it can be visually recognized accurately if there is no defect in the character. The part is identified as defective and is determined to be defective. Therefore, by increasing the line width and the length of the character and giving the allowable range of picture misalignment, it can be determined that the printing is correctly performed if only a slight misalignment occurs. On the other hand, by increasing the line width and length of the reference character for determination, it is not possible to detect missing points or shortened points that are inside the reference character, or reduce the number of missing pixels. May be detected.

  Therefore, the present invention can detect a defect of a minute point existing apart from a line of a character or the like, and a defect in which a line of the character or the like is shortened in an image including a character and a pattern to be inspected. An object is to provide an image loss detection device and an image loss detection method thereof.

Image loss detection apparatus embodiment according to the present invention sets the first picture misalignment tolerance for the reference characters and the reference position deviation of the pattern of the inspection object that is part of the reference image, the first pattern deviation tolerance A first comparison unit that compares the reference image in which the reference image is set with an inspection image including the acquired inspection target, and a second pattern displacement allowable range with respect to the positional displacement of the character and the pattern of the inspection target included in the inspection image. set, and said inspection image second pattern deviation allowable range is set, and the second comparison unit for comparing the reference image including the inspection target, from the first comparing portion and the second comparator unit In each of the output comparison results, when one of the comparison results includes a comparison result indicating that the image to be inspected has a defect, it is determined that there is an image defect and a defect is determined. And a determination unit.

Furthermore, the image loss detection method of the embodiment converts the reference character and the reference pattern to be inspected included in the reference image into an image pattern having a thick or thin line width by changing the density threshold. generates a reference comparison reference image of the first comparison is set to the first picture misalignment tolerance for even if positional deviation inspection target accurately determining a reference image for the comparison, the acquired inspection An image in which the line width is thickened or thinned by changing the density threshold value for the first comparison for comparing with the inspection image including the target and for the characters and patterns to be inspected included in the acquired inspection image The inspection image for comparison is generated by converting the pattern into a pattern and setting the second allowable range of pattern deviation for accurately determining even if the inspection object is displaced. And the reference character and the base Second comparison comparing the reference image including a pattern, of the respective comparison result output from the second comparison between the first comparison, the at least one comparison result, defects in the inspection image And a determination that the inspection image is determined to be defective when the occurrence of the error occurs.

  According to the present invention, in an image including a character and a pattern to be inspected, it is possible to detect a defect of a minute point existing apart from a line of a character or the like and a defect in which the line of the character or the like is cut and shortened. It is possible to provide an image loss detecting device and an image loss detecting method thereof.

FIG. 1 is a diagram conceptually illustrating a configuration example of an image loss detection device according to an embodiment. FIG. 2 is a diagram illustrating a configuration example of the first comparison unit. FIG. 3 is a diagram illustrating a configuration example of the second comparison unit. FIG. 4A is a diagram for explaining detection of an image defect for independent points. FIG. 4B is a diagram for explaining detection of an image loss with respect to an independent point, following FIG. 4A. FIG. 5 is a flowchart for explaining the detection of an image loss by the image loss detection device of the embodiment. FIG. 6A is a diagram for explaining detection of an image defect for a point surrounded by the periphery. FIG. 6B is a diagram for explaining detection of an image loss for a point surrounded by the periphery, following FIG. 6A. FIG. 7A is a diagram for explaining detection of an image loss for a shortened line. FIG. 7B is a diagram for explaining the detection of the image loss with respect to the shortened line, following FIG. 7A.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
The image loss detection device according to the embodiment, with respect to the reference character and the reference pattern to be inspected included in the predetermined reference image, an image pattern having a thick or thin line width by changing the density threshold. By converting, a reference image for comparison is created that is a reference for the first comparison in which an allowable range of picture displacement for accurately determining even if the printed inspection object is displaced is generated, and the reference image for comparison is generated. The first comparison unit that compares the acquired inspection image with the inspection image, and the characters and patterns to be inspected included in the acquired inspection image have a thicker line width by changing the density threshold or The comparison inspection image of the second comparison standard is created by converting the image pattern into a thin image pattern and setting the allowable range of the pattern deviation for accurately determining the positional deviation of the printed inspection object, and detecting the comparison. Image and reference characters Comprising a second comparator for comparing the reference image including the reference picture or the like, and in addition to the normal flash or defect, defects also detected to fine points and short lines of, implementing accurate good or bad determination.

FIG. 1 is a diagram conceptually illustrating a configuration example of an image loss detection device according to an embodiment.
The image loss detection device 1 includes an image pickup unit 2, an image loss detection unit 3, a display unit 4, and an input unit 5.
In the present embodiment, the imaging unit 2 optically reads printed characters, symbols, and patterns, and generates image data (digital signal) of a still image by photoelectric conversion. The image capturing unit 2 captures a still image in a frame unit by a so-called digital camera using a CCD image sensor or a CMOS image sensor, or a line sensor in which the image sensors are arranged in a line. The image pickup unit 2 using the line sensor moves (including parallel movement and rotation movement) the line sensor itself or the inspection object and takes an image by scanning. The imaging unit 2 using the line sensor is suitable for reading characters on a label attached to a curved surface.

  The display unit 4 uses, for example, a liquid crystal display monitor. An operation screen, an inspection status, and a determination result are displayed on the display unit 4 under the control of the control unit 14. The input unit 5 uses, for example, a combination of various switches, an input device such as a keyboard and a mouse. Further, as the input unit 5, a touch panel may be provided on the screen of the display unit 4 and used.

  The image loss detection unit 3 includes an image processing unit 11, a comparison determination unit 12, a reference image data storage unit 13, and a control unit 14. The image loss detection unit 3 can also be configured with a general-purpose personal computer by using a dedicated program or application software.

  The image processing unit 11 uses a known image processing method, as shown in FIGS. 4A and 4B described later, to divide the image data output from the imaging unit 2 into pixels by sampling, and to determine the density of each pixel. (Brightness) is digitized into a brightness value (gray scale) of 256 gradations by quantization. In the present embodiment, since it suffices to determine the presence or absence of an image loss, it is digitized into two gradations of two black and white colors (255: maximum value, 0: minimum value). Of course, the gray scale is not limited to two gradations, and may be a gray scale having a plurality of gradations. Further, in the case of loss detection by a color image, a scale with a plurality of gradations may be set for each of the three colors.

The reference image data storage unit 13 stores a reference image including a non-defective inspection object that is a determination reference captured by the image capturing unit 2. Here, the stored image data is, for example, a quantized reference image of two gradations.
The control unit 14 controls the entire image loss detection unit 3. The instruction to the control unit 14 is given through the input unit 5, and the dedicated program or application software to be installed is executed, and the loss detection and the determination described later are performed.

Here, FIGS. 4A and 4B will be described.
4A and 4B show a partial image of the reference image and the inspection image that includes the same inspection target. In the following description, an image including a character or a pattern to be inspected as a determination reference is referred to as a reference image or a reference image pattern, and an image including a character or a pattern to be inspected is referred to as an inspection image or an inspection image pattern. There is. In addition, the density (color density) and the lightness (brightness) of the two-gradation pixel are used in the same meaning.

  The reference image pattern 4A1 shown in FIG. 4A is a reference image pattern in which the reference image including the reference dots D1 (.) To be detected is sampled and divided into black and white for each pixel. Here, when xy coordinates are set for the pixels in the reference image pattern 4A1, 12 pixels of pixels x3y4, x3y5, x4y3, x4y4, x4y5, x4y6, x5y3, x5y4, x5y6, x6y4 and x6y5 are black, and the reference dot D1 is Is forming. Other pixels around the reference dot D1 are shown as white. Further, the inspection image pattern 4A2 shown in FIG. 4A is an imaged inspection image, and due to an image defect, there is no inspection dot D1 (indicated by a thick frame line) that should originally exist, and all pixels are white. There is. The image processing unit 11 performs these sampling processes. The reference dot D1 and the inspection dot D1 have the same size.

  The reference image pattern 4B1 shown in FIG. 4B is a reference image pattern in which the sampled reference image is quantized and digitized with a density (0, 255) of two gradations. Further, the inspection image pattern 4B2 is an inspection image pattern in which the sampled inspection image is quantized and digitized with a density of one gradation (255). The quantization processing is performed by the image processing unit 11.

  In the reference image pattern 4B3, the first comparing unit 15 described later gives a pixel deviation allowable range to the pixels forming the reference dot D1 and extends the range of pixels to be compared in order to detect the imaged inspection dot D1. 7 shows a pixel pattern (determination area DRAO) with a large value. The allowable range of the pattern deviation can be changed by changing the threshold value (density 0 to 255) of the pixel.

  The allowable range of picture displacement is the allowable range of misregistration with respect to the printing positions of characters and pictures. Specifically, when the printed inspection dots D1 are slightly displaced, if the inspection dots D1 are detected within the set allowable displacement range, the pattern displacement allowable range is normally printed. It is recognized that it has been performed and is judged as good. When the allowable range of the pattern deviation is not set, even if the inspection dot D1 is slightly misaligned and printed, it is determined that a part of the image is missing, and it is determined as a defect. In other words, a slight misalignment of the printed characters is within the permissible range because the information is accurately transmitted when visually recognized by a person, but exceeds the mechanical permissible range, and is therefore defective. The allowable range of the pattern deviation allows the mechanical judgment to have a range of the allowable range.

  The inspection image pattern 4B4 is an inspection image used for comparison by the second comparison unit 16 which will be described later, and is an inspection image pattern in which all pixels are digitized with a density of one gradation (255). Here, the black inspection dot D1 does not exist due to the image defect. The comparison result image pattern 4B5 shown in FIG. 4B is a diagram showing the comparison result of the first comparison unit 15, and the comparison result image pattern 4B6 is a diagram showing the comparison result of the second comparison unit 16.

The comparison / determination unit 12 includes a first comparison unit 15, a second comparison unit 16, and a determination unit 17. The first comparison unit 15 will be described with reference to FIG.
The first comparison unit 15 includes a rank filter 31, a threshold value increase conversion unit 32A, a threshold value decrease conversion unit 32B, threshold value storage units 33A and 33B, and comparison units 34A and 34B. The first comparing unit 15 sets a setting range of the pattern deviation allowable value for the detection target (here, the dot D1) included in the reference image, compares it with the detection target of the inspection image, and makes a comparison for determination. Output the result.

  In the present embodiment, the rank filter 31 sets the number of pixels to be increased outward or the number of pixels to be decreased inward in the vicinity of the pixel of interest, thereby setting the setting range of the allowable pattern deviation value described later. .. The pixel of interest is, for example, a pixel located at the contour of a pixel group forming a character or a picture. In addition, reducing the number of pixels to the inner side means cutting inward by an arbitrary number of pixels from the outermost pixels in the pixel group forming the inspection target. For example, in the case of a character, it means that the character width becomes narrower and thinner. The threshold value is converted by the threshold value increasing conversion unit 32A and the threshold value decreasing conversion unit 32B in the subsequent stage with respect to the set number of pixels. For example, when one pixel is set outside the pixel of interest x3y4, eight pixels x2y3, x2y4, x2y5, x3y3, x3y5, x4y3, x4y4, x4y5 are targeted.

  Further, the rank filter 31 sets the density (brightness) of the brightest pixel among the pixels around the pixel of interest as the maximum rank DRAI, and conversely, the density of the darkest pixel among the surrounding pixels. (Brightness) is set as the minimum rank DRII. In the present embodiment, since the purpose is to detect the presence or absence of image loss, the threshold value for determination may be binary, and the density of pixels in the reference image is binary. Here, the maximum rank DRAI is set to 255 (white) and the minimum rank DRII is set to 0 (black). In a reference image pattern 4B1 of FIG. 4B, which will be described later, for example, if the pixel of interest is pixel x3y4, eight surrounding pixels x2y3, x2y4, x2y5, x3y3 have the maximum rank DRAI: 255 (white), and pixel x3y5, x4y3, x4y4, and x4y5 are the minimum rank DRII: 0 (black).

In the present embodiment, the rank filter 31 converts the threshold value so that the setting range of the allowable pattern deviation value for the inspection target becomes large.
The threshold increase conversion unit 32A determines the inspection target (dot D1 in this case) as the threshold range from the minimum rank DRII to the maximum rank DRAI for the eight pixels around the pixel of interest set by the rank filter 31. ) The range of the concentration threshold is increased (increased). That is, the density maximum / minimum range serving as the threshold value is increased from 0 to 255, and is used as a reference for quality determination in the reference image.
The range of the density threshold value is not limited, and in the present embodiment, the upper and lower limits may be set as two values in order to facilitate the explanation, but it is not necessary to set the maximum and minimum ranks. It does not have to exist and may be between 0 and 255. Further, the density threshold may be set in plural, that is, an intermediate value may be set. This is because when there are gray-scale characters or patterns in the middle gradation in the inspection image, the numerical values of the acquired image data will vary depending on the irradiation state of the illumination light for inspection and the ink density of the characters. Occurs. Further, in the case of characters, due to the delicate positional relationship between the image sensor and the inspection object, it is considered as a mixture of black ink and a white background portion, and has an intermediate value. For example, even in the case of two-color printing of white and black, a value between 0 and 255 (for example, 128) is acquired for the boundary portion. At that time, since the threshold value takes a halfway value such as (128, 255) only with the rank filter 31, the threshold value is converted by the following calculation formula in order to sufficiently expand the value. Here, the comparison reference image DRAO output by the threshold increase conversion unit 32A has a relationship of DRAO = DRAI × AN + BN + (DRAI−DRII) × CN. The parameters of AN, BN, and CN are set arbitrarily, and there are also upper limit threshold values (for increasing values) and lower limit threshold values (for decreasing values), and different threshold conversions are performed. ..

  A reference image DRAO for comparison is set with the determination region D1B in which the upper pixel of the pixel shown in the reference image pattern 4B3 of FIG. 4B is 255 and the lower pixel is 0. The comparison reference image DRAO is stored in the threshold storage unit 33A.

  On the contrary, the threshold reduction conversion unit 32B reduces the number of pixels inward around the pixel of interest based on the number of pixels set by the rank filter 31. By changing the minimum rank DRII to the maximum rank DRAI for the pixels around the pixel of interest, the range of the pattern deviation allowable amount is reduced (reduced) so that the size of the reference dot D1 (black) is reduced. ). That is, the threshold density is converted from 0 to 225. This threshold-converted comparison reference image DRIo is stored in the threshold storage unit 33B. Here, the comparison reference image DRIO output by the threshold reduction conversion unit 32B has a relationship of DRIO = DRII × DN + EN + (DRAI−DRII) × FN. The parameters of DN, EN, FN are set arbitrarily. The threshold storage unit 33A and the threshold storage unit 33B are provided to match the timing of calculating the comparison reference image DRAO and the timing of outputting the inspection image from the image processing unit 11.

The comparison unit 34A has a comparison reference image THMA (equivalent to the reference image DRAO), which is the reference image pattern 4B3 shown in FIG. 4B read from the threshold value storage unit 33A, and the quantized inspection image pattern shown in FIG. 4B. The inspection image of 4B2 is compared in pixel units. In this comparison, a comparison result is obtained in which the inspection image that is the inspection image pattern 4B2 matches the comparison reference image THMA (or is included in the setting range of the pattern deviation allowable value). As will be described later, this comparison result shows the comparison result of the good judgment although the inspection dot D1 is actually missing. The comparison unit 34A sets the comparison result DETNA to 1 when the inspection image DI> the comparison reference image THMA. Further, the comparison unit 34B sets the comparison result DETNI to 1 when the inspection image DI <the comparison reference image THNI.
The output unit 35 is composed of an OR circuit and outputs 1 as the comparison result when either or both of the comparison result DETNA and the comparison result DETNI are 1.

A configuration example of the second comparison unit will be described with reference to FIG.
The second comparison unit 16 includes a rank filter 41, a threshold increase conversion unit 42A, a threshold decrease conversion unit 42B, and comparison units 43A and 43B. The second comparison unit 16 sets the setting range of the allowable pattern deviation value for the detection target included in the inspection image, compares the setting range with the detection target included in the reference image, and outputs the determination comparison result. In the following description, the components that perform the same operations as the components of the first comparison unit 15 described above simplify the description.

  The rank filter 41 has the same function as the rank filter 31 described above, and sets the density (brightness) of the brightest pixel among the pixels around the pixel of interest as the maximum rank DDAI, and on the contrary, exists in the surroundings. The density (brightness) of the darkest pixel among the pixels is set as the minimum rank DDII. In the rank filter 41, the number of pixels to be increased outside in the vicinity of the target pixel is set on the basis of the preset image shift tolerance value.

  The threshold increase conversion unit 42A is equivalent to the threshold increase conversion unit 32A described above, and enlarges the setting range of the allowable pattern deviation value for performing the determination on the inspection dot D1 that should be included in the inspection image. However, in the image processing unit 11, since the inspection dot D1 is missing in the image pattern 4A2 shown in FIG. 4A, when the quantization processing is performed, the density of all pixels is 255 as shown in the inspection image pattern 4B2. (White). When the inspection dot D1 is present, the threshold increase conversion unit 42A sets the determination area D1B as shown in the reference image pattern 4B3. Here, the comparison reference image DDAO output by the threshold increase conversion unit 42A has a relationship of DDAO = DDAI × AI + BI + (DDAI−DDII) × CI for each rank. The parameters of AI, BI, and CI are set arbitrarily. However, since the inspection dot D1 is missing, actually, when the maximum rank DDAI of the inspection image pattern 4B2 is set to 255 (white), as shown in the inspection image pattern 4B4, all pixels having allowable values of 255 are white. Is set.

  Also in the threshold reduction conversion unit 42B, since the inspection dot D1 does not exist in the image pattern 4A2 which is the inspection image and the density is only 255, even in the inspection image pattern 4B4, the allowable pattern deviation values are all pixel values. The density is set to 255. Here, the comparison reference image DDIO output by the threshold reduction conversion unit 42B has a relationship of DDIO = DDII × DI + EI + (DDAI−DDII) × FI for each rank. The parameters of DI, EI and FI are set arbitrarily.

  The comparison units 43A and 43B both compare the inspection image pattern 4B4 and the reference image pattern 4A1 for each pixel, and thereby a comparison result image pattern 4B6 including the missing dot D1 is output. That is, it is detected that the dot D1 is missing. The comparison unit 43A sets the comparison result DETIA to 1 when the reference image REM> the reference image for comparison DDAO. Further, the comparison unit 43B sets the comparison result DETII to 1 when the reference image REM <the comparison reference image DDIO.

The output unit 44 is composed of an OR circuit, and outputs 1 as the comparison result when either or both of the comparison result DETIA and the comparison result DETII are 1.
To the determination unit 17, the comparison result without image loss is input from the first comparison unit 15, and the comparison result with image loss is input from the second comparison unit 16. That is, the determination unit 17 determines that there is an image loss when one of the comparison results includes 1, and determines the defect.

Hereinafter, with reference to the flowchart shown in FIG. 5, the image defect inspection by the image defect detection device 1 of the present embodiment will be described.
First, according to an instruction from the control unit 14, the reference image data storage unit 13 stores a reference image REM including a non-defective inspection target which is a comparison reference captured by the image capturing unit 2 or already converted into image data (step S1). ). The reference image REM stored in the reference image data storage unit 13 is digitized by the image processing unit 11 and divided into pixels by sampling to quantize the density of each pixel, and the reference image pattern 4B1 shown in FIG. 4B. May be At this time, the first comparison unit 15 generates the comparison reference images DRAO and DRIO including the reference image pattern 4B3 in which the threshold value is converted with respect to the reference image pattern 4B1 and the setting range of the allowable pattern deviation value is set. It may be stored in the threshold storage units 33A and 33B.

  Next, the image capturing section 2 captures the inspection image DI including the dots D1 to be detected (step S2). The image processing unit 11 digitizes the inspection image DI captured by the imaging unit, divides it into pixels by sampling, and generates the inspection image pattern 4B2 in FIG. 4B in which the density of each pixel is quantized (step S3). ).

  First, the first comparison unit 15 reads the reference image REM from the reference image data storage unit 13 and changes the threshold value of the reference image pattern 4B1 to set the setting range of the allowable pattern deviation value. , DRIO are generated and stored in the threshold storage units 33A and 33B (step S4). Of course, as described above, when the reference image REM is input to the image loss detection device 1, the reference images DRAO and DRIO for comparison may be generated and stored in the threshold value storage units 33A and 33B.

  Further, the first comparison unit 15 reads the comparison reference images THMA and THMI from the threshold storage units 33A and 33B to the comparison units 34A and 34B. In the comparison units 34A and 34B, the comparison reference images HMA and THMI (reference image pattern 4B3) and the inspection image (inspection image pattern 4B2) are compared for each pixel, and comparison results DETNA and DETNI are output (step S5). .. That is, the comparison unit 34A compares the comparison reference image THMA (equivalent to the reference image DRAO) read from the threshold value storage unit 33A and the quantized inspection image pattern 4B2 shown in FIG. 4B in pixel units. ..

  The comparison unit 34B compares, for each pixel, the comparison reference image THMI (equivalent to the reference image DRIO) read from the threshold storage unit 33B and the quantized inspection image pattern 4B2 illustrated in FIG. 4B. The reference images DRAO and DRIO for comparison are used for writing and the reference images THMA, THMI for comparison are used for reading, but they are equivalent.

  In these comparison results DETNA and DETNI, the inspection image pattern 4B2 coincides with or is included in the comparison reference image THMA, so that the comparison result of the comparison result image pattern 4B5 is obtained. As a result of this comparison, since the inspection dot D1 cannot be extracted in spite of the fact that the inspection dot D1 is defective, the comparison result is output as a comparison result that is determined as good. Of course, such minute dots (dots) distant from a character or the like cannot be detected, but a large loss of ordinary characters, so-called character loss, stains, and bleeding are detected.

  Simultaneously with the processing in the first comparison unit 15, the second comparison unit 16 causes the threshold increase conversion unit 42A to set the inspection image pattern 4B4 for setting the pattern deviation allowable range with respect to the inspection image pattern 4B2 shown in FIG. 4B. A comparison inspection image DDAO is generated and output to the comparison unit 43A (step S6). Similarly, the threshold reduction conversion unit 42B generates a comparison inspection image DDIO of an image pattern for setting the allowable range of the pattern deviation for the inspection image pattern 4B2 shown in FIG. 4B and outputs it to the comparison unit 43B.

  Next, the comparison units 43A and 43B both compare the comparison inspection image (inspection image pattern 4B4) and the reference image (reference image pattern 4A1) for each pixel (step S7). By this comparison, the comparison result of the comparison result image pattern 4B6 shown in FIG. 4B including the defective inspection dot D1 is output.

  The determination unit 17 determines that the defect determination result is included from the comparison result of the good determination from the first comparison unit 15 and the comparison result of the defect determination from the second comparison unit 16 (step S8). The determination unit 17 outputs the comparison result of the defect determination from the second comparison unit 16 to the control unit 14 as the determination result. The control unit 14 causes the display unit 4 to display the comparison result image pattern 4B6 to display that the dot D1 is missing in the inspection image DI (step S9).

  As described above, according to the image loss detecting apparatus of the present embodiment, the allowable pattern deviation range is set for the reference character and the reference pattern to be inspected included in the reference image, and the inspection image including the inspection target is set. In addition to the first comparison for comparing with, the allowable range of pattern deviation is set for the characters and patterns to be inspected included in the inspection image, and the second comparison for comparing with the reference image is performed. According to the comparison result obtained by combining the first comparison and the second comparison, in addition to the detection of normal defects and extra print stains adhering to the blank portion, the defects of minute points distant from the character or the pattern are detected. Can be detected, and accurate pass / fail determination can be performed.

Next, with reference to FIGS. 6A and 6B, as a first example of loss by the image loss detection apparatus 1 of the above-described embodiment, detection of an image loss with respect to a point surrounded by the periphery will be described.
In the first example, the reference image shown in FIG. 6A is a reference image in which a minute point 53 is printed on the inside of the cup-shaped frame line 52, is sampled, and is divided into black and white for each pixel. This is the image pattern 6A1. Further, the inspection image is the reference image pattern 6A2 in which only the frame line 52 remains and the image in which the point 53 is missing is subjected to sampling processing and is divided into black and white for each pixel. Here, when the xy coordinates are set for the pixels in the reference image pattern, the frame line 52 is formed in a cap shape by the pixels x1y1 to x1Y9 on the left side, the pixels X2y9 to X9y9 on the upper side, and the pixels x10y1 to x10y9 on the right side. It The point 53 is formed by 18 pixels of pixels x3y5, x4y4 to x4y6, x5y3 to x5y7, x6y2 to x6y6, x7y3 to x7y5, x8y4.

  An image pattern 6B1 shown in FIG. 6B is a reference image pattern in which the sampled reference image is quantized and digitized with a density (0, 255) of two gradations. Similarly, the inspection image pattern 6B2 is an inspection image pattern in which the sampled inspection image is quantized and digitized with a density (0,255) of two gradations. The quantization processing is performed by the image processing unit 11.

  In the reference image pattern 6B3, the first comparison unit 15 gives the above-described pixels forming the frame line 52 and the point 53 an allowable range of the pattern deviation (density 0 to 255), and the frame line of the imaged inspection image pattern. A pixel pattern in which pixels for detecting 52 and the point 53 are enlarged outside is shown. Also in this example, similarly to the above, the rank filter 31 is set to increase one pixel to the outside or decrease one pixel to the inside in the periphery adjacent to the pixel of interest. Therefore, the eight pixels around the target pixel become the setting range of the allowable pattern deviation value.

  When the setting range of the allowable pattern deviation value is increased by one pixel to the outside, for example, if the pixel of interest is x3y5, the surrounding pixels x2y4, x2y5, x2y6, x3y4, x3y6, x4y4, x4y5, x4y6 are 8 pixels. Is the target. These eight pixels are set in the range of the allowable pattern deviation value (density 0 to 255) by the maximum rank 255 and the minimum rank 0. Therefore, in the reference image pattern 6B3 shown in FIG. 6B, most of the pixels other than a few pixels are set within the range of the allowable pattern deviation value (density 0 to 255).

  The first comparison unit 15 compares the comparison reference images THMA (comparison reference image pattern 6B3) and THMI with the inspection image (inspection image pattern 6B2) for each pixel, and compares the comparison results DETNA and DETNI (comparison result image pattern). 6B5) is output. In the comparison result image pattern 6B5, since the reference image pattern 6B3 includes the inspection image pattern 4B2 except for the two pixels, most of the pixels at the point 53 are missing and only two pixels remain. Image pattern.

  In addition, the second comparison unit 16 sets the pattern deviation allowable range for the inspection image pattern 6B2 to generate the comparison inspection images DDAO (comparative inspection image pattern 6B4) and DDIO. The second comparison unit 16 compares the comparison inspection images DDAO and DDIO with the reference image (reference image pattern 6B1) for each pixel. In the comparison between the comparison inspection image pattern 6B4 and the reference image pattern 4A1, the comparison result of the comparison result image pattern 6B6 is output.

  In the first example, when the allowable pattern deviation range is set for the pixel of the inspection target point 53 included in the reference image, the pattern deviation occurs due to contact with or overlapping with the pixel in the allowable pattern deviation range of the frame line 52. The allowable range is too wide, and the detection of the inspection target point 53 included in the inspection image is likely to be uncertain.

    According to the first example, when the point 53 to be inspected, which is supposed to be included in the inspection image, is missing, the point 53 is not affected by the pixels in the allowable range of the pattern deviation of the frame line 52. The missing point 53 can be detected when compared with the existing reference image pattern. According to the comparison result obtained by combining the first comparison and the second comparison, it is possible to detect the defect of the point surrounded by the periphery in addition to the detection of the normal defect and the extra print stain adhered to the blank portion. Therefore, it is possible to perform accurate pass / fail determination.

Next, with reference to FIGS. 7A and 7B, as a second case of loss by the image loss detection apparatus 1 of the above-described embodiment, detection of an image loss having a shortened length will be described.
In the second example, the reference image shown in FIG. 7A is the reference image pattern 7A1 in which the image pattern including the end portion 61 such as a character is sampled and is divided into black and white for each pixel.
Further, the inspection image is an inspection image pattern 7A2 in which each image is divided into black and white for each pixel by subjecting the image pattern, which is shortened due to the loss of the front end portion 63 of the end portion 62 such as the discontinuity of characters, to a sampling process.

Here, when the xy coordinates are set for the pixels in the reference image pattern, the end portion 62 is formed by the pixels x2y5, X3y4 to X3y6, x4y3 to x4y7, x5y2 to x5y9, x6y3 to X6y8, x7y4 to x7y8, x8y5 to x8y8. There is. In the inspection image pattern 7A2, the pixels x2y5, X3y4 to X3y6, x4y3 to x4y6, x5y2 to x5y5, x6y3, and X6y4 forming the tip portion 63 from the end portion 62 are defective. The image pattern 7B1 shown in FIG. It is a reference image pattern that has been quantized and digitized with a density of two tones (0,255). Similarly, the inspection image pattern 7B2 is an inspection image pattern in which the inspection image is quantized and digitized at a density of two gradations (0,255). The quantization processing is performed by the image processing unit 11.

  In the reference image pattern 7B3, the first comparison unit 15 gives the above-described pixels forming the end portion 61 a picture pattern deviation allowable range (density 0 to 255) to detect the end portion 61 of the imaged inspection image pattern. The pixel pattern in which the range for enlarging the pixel for doing is enlarged is shown. In this case as well, the rank filter 31 is set to increase one pixel outward or decrease one pixel inward in the periphery adjacent to the pixel of interest. Therefore, the eight pixels around the target pixel become the setting range of the allowable pattern deviation value.

When the setting range of the allowable pattern deviation value is increased by one pixel outward, for example, if the pixel of interest is x2y5, the surrounding pixels x1y4, x1y5, x1y6, x2y4, x2y6, x3y4, x3y5, x3y6 are 8 pixels. Is the target. These eight pixels are set in the range of the allowable pattern deviation value (density 0 to 255) by the maximum rank 255 and the minimum rank 0. Therefore, in the reference image pattern 7B3 shown in FIG. 7B, most of the pixels other than the pixels at the three corners are set within the range of the allowable pattern deviation value (density 0 to 255).
The first comparing unit 15 compares the comparison reference images THMA (comparison reference image pattern 7B3) and THMI with the inspection image (inspection image pattern 7B2) pixel by pixel, and compares the comparison results DETNA and DETNI (comparison result image pattern). 7B5) is output.

  Regarding these comparison results DETNA and DETNI, since the inspection image pattern 7B2 is included in the comparison reference image THMA, the comparison result of the comparison result image pattern 7B6 is obtained. The result of this comparison is an image pattern in which most of the pixels are lost and only three pixels remain, though the end portion 62 is the end portion 62 where the tip portion 63 is missing, in reality. .. Therefore, the difference between the reference image and the inspection image is small, and the comparison result by the determination with low accuracy is output.

The second comparison unit 16 sets the pattern deviation allowable range for the inspection image pattern 7B2 to generate the comparison inspection images DDAO (comparative inspection image pattern 7B4) and DDIO. The second comparison unit 16 compares the comparison inspection images DDAO and DDIO with the reference image (reference image pattern 6B1) for each pixel. In the comparison between the comparison inspection image pattern 7B4 and the reference image pattern 7A1, the comparison result of the comparison result image pattern 7B6 is output.
In the comparison result image pattern 7B6, the end portion 62 lacking the tip end portion 63 is shorter than the end portion 61 included in the reference image even if the end portion 62 includes the range of the allowable pattern deviation value. Since there is a difference between them, a defect can be detected.

  According to the second example, even if a part of the end portion such as the discontinuity of characters is lost by setting the range of the pattern deviation allowable value for the inspection object having the defect included in the inspection image. Since it is not affected by the pixels in the allowable range of the pattern deviation, the defective end portion 62 can be detected when compared with the reference image pattern. According to the comparison result obtained by combining the first comparison and the second comparison, in addition to the detection of normal defects and extra print stains adhering to the blank portion, the defects of minute points distant from the character or the pattern are detected. Can be detected, and accurate pass / fail determination can be performed.

1 ... Image loss detection device, 2 ... Imaging unit, 3 ... Image loss detection unit, 4 ... Display unit, 4A1, 4B1, 4B3 ... Reference image pattern, 4A2, 4B2, 4B4 ... Inspection image pattern, 4B5, 4B6 ... Comparison result Image pattern, 5 ... Input unit, 11 ... Image processing unit, 12 ... Comparison determination unit, 13 ... Reference image data storage unit, 14 ... Control unit, 15, 16, 34A, 34B ... Comparison unit, 17 ... Determination unit, 31 , 41 ... Rank filter, 32A, 42A ... Threshold increase conversion unit, 32B, 42B ... Threshold decrease conversion unit, 33A, 33B ... Threshold storage unit, 43A, 43B ... Comparison unit, D1 ... Dot (reference dot, inspection dot).

Claims (4)

Set the first picture misalignment tolerance to inspection standard characters and the reference position deviation of the pattern of the object included in the reference image, including a said first pattern deviation tolerances set reference image, a test object acquired A first comparing section for comparing the inspection image,
Set the second pattern deviation tolerance for the character and the positional deviation of the pattern of said object included in the inspection image, the inspection image second pattern deviation tolerance is set, the including the inspection target A second comparing section for comparing with the reference image;
In each of the comparison results output from the first comparison unit and the second comparison unit, when one of the comparison results includes a comparison result indicating that the image to be inspected has a defect, A determination unit that determines that there is a defect and determines a defect,
An image loss detection device comprising: An image processing unit that performs sampling processing and quantization processing on the reference image and the inspection image,
Oite the first comparator unit,
By setting the numerical value of the density with respect to the pixels around the pixels constituting the inspection target whose density is digitized by the image processing unit, the first pattern deviation allowable range is set ,
In the second comparison unit,
The second allowable deviation range is set by changing the numerical value of the density with respect to the pixels around the pixel constituting the inspection target whose density is digitized by the image processing unit. 1. The image defect detection device according to 1. The first comparison unit is
At least two thresholds of different sizes are set for the densities digitized by the image processing unit,
A threshold value increase conversion unit that sets the first allowable pattern deviation range using the larger threshold value for pixels arranged on the outer periphery of the pixel group forming the inspection target;
A threshold value reduction conversion unit that sets the first allowable pattern deviation range using the smaller threshold value for the pixels arranged on the outer periphery of the pixel group forming the inspection target;
The second comparison unit is
At least two thresholds of different sizes are set for the densities digitized by the image processing unit ,
A threshold value increase conversion unit that sets the second allowable pattern deviation range using the larger threshold value for pixels arranged on the outer periphery of the pixel group forming the inspection target;
A threshold value decrease conversion unit that sets the second allowable pattern deviation range by using the smaller threshold value for pixels arranged on the outer periphery of the pixel group forming the inspection target;
The image loss detection device according to claim 2, further comprising: For the reference character and reference pattern to be inspected included in the reference image, the density threshold is changed to convert it to an image pattern with a thick or thin line width, and even if the inspection target is misaligned, it is accurate. First, a comparison reference image of a first comparison reference in which a first pattern displacement allowable range for making a determination is set is generated, and the comparison reference image and the acquired inspection image including the inspection target are compared. Comparison of
For the characters and patterns to be inspected included in the acquired inspection image, the density threshold is changed to convert the image pattern into a thicker or thinner line width, and even if the inspection target is misaligned. generates reference comparative inspection image of the second comparator is set to the second pattern deviation tolerance for accurately determining, with said comparative test image and the reference image including the reference characters and the reference picture A second comparison to compare,
Of the respective comparison results output from the first comparison and the second comparison, when at least one comparison result has a defect in the inspection image, the inspection image is determined to be defective. Judgment,
An image loss detection method comprising: