JP2011076204A - Method and apparatus for inspecting printed matter - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a highly accurate and high-speed printed matter inspecting method and a printed matter inspecting apparatus for performing positional correction relating to uneven geometric distortion within an inspected image caused by displacement, expansion, contraction, etc. of a printed matter. <P>SOLUTION: In the printed matter inspecting method, a periphery of a detected defect candidate is searched after performing first positional correction using a standard image and the inspected image, and whether it is a real defect is determined by performing second positional correction relating to the defect candidate determined to have high possibility of error detection caused by the geometric distortion. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a printed matter inspection method and a printed matter inspection apparatus for detecting defects in printed matter (letter missing, blurring, dirt, etc.).

  Conventionally, an inspection for determining the quality of a printed material is performed by comparing an image (inspection image) obtained from a printed material to be inspected with reference to an image (reference image) obtained from a printed material that has been determined as a good product by visual inspection or the like in advance. Is done. In this case, in the comparison process, the reference image and the inspection image are aligned, and a luminance difference is calculated for each pixel. In the case of a normal pixel, the difference value becomes very small, but in the pixel corresponding to a defective portion such as missing characters or ink splashes, the difference value becomes large. Therefore, by binarizing with a preset threshold value, only a defective portion can be extracted as a feature point.

  However, since printed materials such as paper and film are soft, they tend to expand and contract, and their occurrence location and frequency are random. Therefore, the inspection image also has a geometric distortion of the pattern, and when the comparison process with the reference image is performed, the difference value becomes large at the outline portion of the pattern, leading to erroneous detection of a defect. Therefore, as a countermeasure against the geometric distortion as described above, a plurality of divided images are created from at least one of the reference image and the inspection image, and the positional deviation of the image for obtaining the positional deviation correction amount for each division unit. A correction method is known (Patent Document 1). In addition, a method is known in which image projection data is calculated for each divided area, a feature amount is searched from the projection data, and an LUT (Look Up Table) relating to a position correction amount is created. However, in the method described in Patent Document 1, when the geometric distortion is unevenly distributed in the inspection image, the geometric distortion remains in the divided image even if the divided image is created. Teshima.

  A method for calculating two-dimensional projection waveform data for a divided image has been proposed (Patent Document 2). However, in order to calculate two-dimensional projection waveform data, the amount of calculation increases, and when high-speed processing is required, a computer having high performance in terms of computer capability is required.

  Furthermore, as another method for preventing erroneous detection, a method has been proposed in which a region where the density of an inspection image changes rapidly is set as a mask region and excluded from the inspection target (Patent Document 3). However, with the method described in Patent Document 3, most of the label or the like on which characters are printed becomes a mask region, and defect detection cannot be performed in regions where characters are missing or density changes.

Japanese Patent No. 3140838 Japanese Patent Laid-Open No. 2004-199548 JP-A-4-339653

  The present invention has been made in view of the above problems, and corrects the position for non-uniform geometric distortion in an inspection image caused by misalignment of printed matter (position misalignment due to meandering of printed matter) or expansion / contraction. An object of the present invention is to provide a method for inspecting a printed matter with high accuracy and high speed and a printed matter inspection apparatus.

As means for solving the above problems, the invention described in claim 1 is a printed matter inspection method for inspecting a printed matter by performing image processing after converting the image of the printed matter into digital image data,
[A] A printed matter that has been determined to be good in advance is captured as a reference image, and image data divided into m × n blocks is created.
[B] Taking an image of a printed matter to be inspected as an inspection image, creating image data divided into m × n blocks,
[C] In a region preset for each block, a pattern search is performed with the block of the reference image corresponding to the block to perform a first position correction process,
[D] A difference process and a labeling process are performed between the inspection image subjected to the first position correction process and the reference image, and defect candidates are extracted.
[E] In the vicinity of the defect candidate, if a defect candidate is a white defect candidate, a black defect candidate is searched, and if the defect candidate is a black defect candidate, a white defect candidate is searched within a preset range. And
[F] In the process [e], when there is a defect candidate corresponding to the defect candidate, the number of pixels corresponding to the line width of the defect candidate is calculated, and the divided image region to which the defect candidate belongs is calculated. A position correction amount is set to the number of pixels, a second position correction process is performed on the inspection image, and a difference process and a labeling process are performed between the inspection image subjected to the second position correction process and the reference image. And
[G] In the process [f], if a defect candidate is detected at the same position as the defect candidate, the defect candidate is determined to be a true defect, and if not detected, the defect candidate is determined to be a false detection. And
[H] In the printed material inspection method, in the process [e], when there is no defect candidate corresponding to the defect candidate, the defect candidate is determined as a true defect.

  The invention according to claim 2 is the printed matter inspection method according to claim 1, wherein the number of m × n blocks is preset according to the degree of ease of expansion and contraction of the printed matter.

  According to a third aspect of the present invention, there is provided a printed matter inspection apparatus for inspecting a printed matter using the printed matter inspection method according to the first or second aspect.

  The printed matter inspection method and the printed matter inspection apparatus of the present invention perform high-precision and high-speed defect correction by performing position correction processing on non-uniform geometric distortion in the inspection image caused by misalignment or expansion / contraction of the printed matter. Can be detected.

Schematic which showed an example of the structure of the printed matter inspection apparatus using the printed matter inspection method which concerns on this invention. Schematic showing the method of pattern matching generally used. The figure for demonstrating a 1st position correction process. (A) is a figure which shows the correspondence of the divided | segmented reference | standard image and test | inspection image. FIG. 6B is a diagram showing a pattern search area around a divided inspection image. The figure which shows the typical example of an image processing result when the geometric distortion generate | occur | produces in the test | inspection image by expansion / contraction of printed matter. (A) is a figure which shows a reference | standard image. (B) is a diagram showing an inspection image. Flow chart of inspection processing in an embodiment of the present invention The figure for demonstrating the method to set position correction amount to this pixel number. (A) is a figure similar to FIG.4 (c), and is a figure which shows the difference image of a reference | standard image and a test | inspection image. FIG. 7B is an enlarged view of a part of FIG. 6A, and is an enlarged view of a left part and a right part of an axis 405 in which an image is stretched in the direction of an arrow. (C) is a diagram for explaining a method of setting the position correction amount to the number of pixels.

  Embodiments of the present invention will be described below with reference to the drawings.

  FIG. 1 is a schematic view showing an example of the configuration of a printed matter inspection apparatus using the printed matter inspection method according to the present invention. The printed material 101 is placed on the roller 102 and is conveyed in the X direction by the rotation of the roller 102. The light irradiation unit 103 is configured, for example, by arranging a plurality of LEDs on the surface. The light irradiation unit 103 has an opening at the center so as not to obstruct the field of view of the imaging unit 104. In the embodiment of the present invention, the light irradiation unit 103 is arranged as described above in order to capture scattered light from the printed material 101. However, depending on the inspection purpose, the arrangement of the light irradiation unit 103 and the light irradiation unit 103 are used. Alternatively, the transmitted light or reflected light from the printed material 101 may be obtained.

  As the imaging unit 104, a camera using a charge coupled device is used. Further, in this embodiment, a line sensor camera in which charge coupled devices are arranged one-dimensionally is used to increase the inspection speed.

  The acquired analog image data is converted into 256-gradation digital image data, and then sent to the image processing unit 105 that performs arithmetic processing, performs data processing, and performs quality determination processing of the printed matter 101.

  A determination process performed using the obtained digital image data will be described.

  FIG. 2 is a schematic diagram showing a generally used pattern matching technique, and is also used for correction of pattern misalignment in the present invention.

  The pattern search method used in the embodiment of the present invention will be described with reference to FIG. The pattern search is a method in which a characteristic image 201 is set in advance and the most similar pattern image is searched from the search target region image 202. A method called template pattern matching is generally used. At this time, the similarity is determined by calculating a normalized correlation coefficient represented by (Equation 1).

ここで、
ｒ：正規化相関係数
Ｎ：画素数
Ｉ：探索対象領域画像２０２内の注目画素の輝度値
Ｍ：特徴のある画像２０１内の注目画素の輝度値

here,
r: normalized correlation coefficient N: number of pixels I: luminance value of the target pixel in the search target region image 202 M: luminance value of the target pixel in the characteristic image 201

The normalized correlation coefficient r takes a value in the range of −1 to 1, and when r = 1, the characteristic image and the pattern image completely match, and when r = 0, the characteristic image and the pattern image are When completely inconsistent and r = −1, it indicates that the characteristic image and the pattern image are inverted images. Within the search target area image 202, the position of the image is the same as the characteristic image while shifting the position by 1 Pix (1 pixel: one pixel of the imaging camera used in the imaging unit) along the direction indicated by the scanning direction 203. An area is set, and Equation 1 is calculated with reference to image data of each pixel. As a result, by searching for the position where the normalized correlation coefficient is closest to 1, the position of the image 204 having the same pattern as the characteristic image 201 can be searched. In the embodiment of the present invention, the pattern search method described above with respect to the position correction process is used.

  The flow of inspection processing in the embodiment of the present invention will be described using FIG. In this case, it is assumed that the reference image has already been set by a printed material determined to be non-defective. After capturing the inspection image (S1), the reference image and the inspection image are divided, and position correction processing is performed for each of the divided images (hereinafter referred to as segments) (S2).

  The position correction process will be described with reference to FIG.

  FIG. 3A is a diagram showing a correspondence relationship between the divided reference image and the inspection image, and FIG. 3B is a diagram showing a pattern search area around the divided inspection image, when performing position correction. Is required.

  First, as shown in FIG. 3A, both the reference image 301 and the inspection image 302 are divided into a total of m × n segments of m in the horizontal direction of the image and n in the vertical direction of the image. In this case, the number of segments to be divided is set in advance according to the ease of expansion and contraction of the printed material, and the inspection accuracy increases by decreasing the number of divisions for a printed material that is easily stretchable and increasing the number of divisions for a printed material that is difficult to expand and contract. . Each segment of the reference image and the inspection image is subjected to the position correction process described below with a pair of those having the same positional relationship as indicated by arrows 304.

  Position correction is performed for each segment of the reference image 301 and the inspection image 302 using the pattern matching method described above. That is, as shown in FIG. 3B, the position correction search range α (image horizontal direction) and β (image vertical direction) set in advance from the original position 306 with respect to one segment in the inspection image. The direction is shifted by 1 Pix (for example, position 307), and the degree of similarity (ie, normalized correlation coefficient) with the corresponding reference image segment is calculated. In this process, the area number in the inspection image is newly set as the inspection image at the segment position with the number of Pixes when shifted to the image having the highest similarity as the position correction amount.

  In the above-described position correction processing, by setting the search ranges α and β to approximately the size of expansion and contraction, it is possible to absorb erroneous detection due to geometric distortion occurring in the inspection image to some extent. The above process is referred to as a first position correction process.

  However, if the degree of positional deviation or expansion / contraction differs in the segment image, it is not possible to prevent erroneous detection only by performing the first position correction process. For example, when the degree of expansion / contraction differs between the right side and the left side in the segment image, it is conceivable that the reference image and the inspection image match well on the right side in the segment image but do not match on the left side, or vice versa. For this reason, a false detection of a defect is likely to occur particularly in the vicinity of the outline of a character.

  In order to prevent the above phenomenon, it is considered that the segment size should be reduced. However, an increase in the processing load due to an increase in the number of segments is expected, and by reducing the segment size, feature points are included in the segment image. There is a problem that the reliability of the similarity is lowered due to the absence of the case. Therefore, in the embodiment of the present invention, in addition to the first position correction process, the second position correction process is performed for a specific defect candidate, thereby enabling high-precision and high-speed defect detection.

  After the first position correction process in step S2, a difference process and a labeling process are performed between the reference image and the inspection image (S3). In the embodiment of the present invention, the brightness value calculated when creating the difference image is calculated as shown in Expression 2.

Brightness value of difference image = Brightness value of inspection image-Brightness value of reference image + 128
... (Formula 2)

  In the difference image created by the above processing, for a pixel having a luminance value of 128 or more, for example, the luminance of the reference image 100 is obtained by performing a labeling process on the pixel having a luminance value of a predetermined threshold value or more. In the case of a white defect having a luminance value of 200 with respect to the value, in Equation 2, the luminance value of the difference image is 228 from 200-100 + 128, and when the threshold is 200, for example, the white defect Candidates are extracted. Further, for example, in the case of a black defect having a luminance value of 100 with respect to the luminance value of the reference image 200, the luminance value of the difference image is 28 from Equation 100, and the luminance value of the difference image is 28. For a pixel having a value, a black defect candidate is extracted by performing a labeling process on a pixel having a luminance value equal to or lower than a predetermined threshold value.

  The presence / absence of a defect candidate is determined by the labeling process (S4), and when a defect candidate exists, the defect mode (a mode such as a black defect or a white defect) is referred to for each defect candidate (S5). If there is no defect candidate, it is determined as a non-defective image, and the inspection is terminated (S14).

  When the defect mode of the defect candidate is a white defect in step S5, a defect candidate whose defect mode is a black defect is searched within a preset search range (S6). Conversely, when the defect mode of the defect candidate is a black defect, a defect candidate with a defect mode of white defect is searched (S7).

  When a defect candidate having a defect mode to be searched in step S6 or step S7 exists around the defect candidate, the defect candidate may be a falsely detected defect candidate due to geometric distortion of the inspection image. It is determined that there is a second position correction process and a difference and labeling process is performed (S8). If the defect candidate having the defect mode to be searched in step S6 or step S7 is not located around the defect candidate, it is determined that the defect candidate is a true defect (S9).

  Here, the meaning of steps S5 to S7 will be described in detail with reference to FIG.

  FIG. 4 is a diagram showing a typical example of the image processing result when geometric distortion occurs in the inspection image due to expansion / contraction of the printed matter, and assumes a case where partial geometric distortion occurs in the inspection image. This is an example. 4A shows a reference image 401 and FIG. 4B shows an inspection image 402. Partial expansion and contraction occurs in the inspection image 402, and the image is expanded in the direction of the arrow about the axis 405 only in the region 404. In the inspection image 402, there are black defects 406 caused by ink smears and white defects 407 caused by missing characters. In the case of the example shown above, the difference image 403 between the reference image 401 and the inspection image 402 has a luminance distribution as shown in FIG. In addition to feature points resulting from true defects, a plurality of white and black defect candidates appear around the region 404 due to geometric distortion of the image. Here, the difference between the defect candidate that appears due to geometric distortion and the true defect candidate is that the defect candidate that appears due to geometric distortion appears as a pair of white and black defects. Defects indicated by 406 and 407 that do not originate are not paired.

Using the above features, pay attention to the periphery of each defect candidate, and when a defect candidate having a defect mode different from the defect mode of the defect candidate exists in a preset search area, the following processing is performed. . That is, for the segment image where the defect candidate is located, the position correction amount is set to the number of pixels of the defect candidate. A method for setting the position correction amount to the number of pixels is shown in FIG. FIG. 6A is a diagram similar to FIG. 4C, and shows a difference image 403 between the reference image 401 and the inspection image 402. FIG. 6B is an enlarged view of a part of FIG. 6A, and is an enlarged view of the left part 501 and the right part 502 of the shaft 405 in which the image is stretched in the direction of the arrow. The black defect candidate and the white defect candidate make a pair. FIG. 6C is a diagram for explaining a method of setting the position correction amount to the number of pixels of the defect candidate. The pixel referred to here is a 1-Pix imaging pixel 408 of the line sensor camera used in the imaging unit 104. The line width of the black defect candidate and the white defect candidate in the case of FIG. 6C corresponds to about two pixels. In the present invention, the position correction amount is set to the number of pixels, that is, two pixels in this case, and the second position correction processing is performed.

  Next, a second position correction processing method will be described. Reference numeral 203 denotes the scanning direction of the line sensor camera. In the case of the defect candidate 501, the black defect candidate is first imaged and then the white defect candidate is imaged. The defect candidate 502 is the image of the white defect candidate. Then, the black defect candidate is imaged. In the second position correction process, when there are white defect candidates next to black defect candidates as in the case of 501 defect candidates in the scanning direction of the line sensor camera, the two pixels are moved backward in the scanning direction. Perform position correction. On the contrary, when a white defect candidate is present next to a black defect candidate as in 502, the position of the two pixels is corrected forward in the operation direction.

  The second position correction process is performed, and the difference and labeling process is performed (S8). It is determined whether or not a defect is detected at the same location as the defect candidate (S10), and if it is detected, it is determined as a true defect site that does not result from expansion and contraction (S11). As a false detection, the defect candidate is determined as a normal part (S12).

  As an example in which the defect modes of the white defect and the black defect appear as a pair, a defect due to blurring of characters can be considered in addition to the image expansion / contraction described above. In this case, defect candidates having the characteristics of the defect mode white defect and black defect appear around the blurred portion, and the second position correction is performed for each defect candidate, and the defect determination is performed again. Since this is a comparison between the raised inspection image and a normal reference image, a blur defect is always detected regardless of the position correction.

  It is determined whether or not Step S5 to Step S12 have been performed for all defect candidates (S13). If it has been performed, the inspection ends, and if it has not been performed, the procedure returns to Step S4.

  According to the printed matter inspection method and the printed matter inspection apparatus of the present invention, it is possible to erroneously determine a defect by performing position correction processing on non-uniform geometric distortion in an inspection image caused by misalignment or expansion / contraction of the printed matter. As a result, defects can be detected with high accuracy and high speed.

101 ... printed matter (inspection object)
DESCRIPTION OF SYMBOLS 102 ... Roller 103 ... Light irradiation part 104 ... Imaging part 105 ... Image processing part 201 ... Reference image 202 ... Inspection image 203 ... Scanning direction 204 ... Pattern search As a result, the region 301 that is determined to have the highest similarity to the reference image... The reference image 302... The inspection image 303.
304: Represents the correspondence between the reference image and the inspection image in the segment image 305 ... The area 306 for position correction ... The original position 307 of the segment image ... The similarity with the reference image is calculated Area of interest 401 ... Example of reference image 402 ... Example of inspection image 403 ... Differential image 404 ... Area 405 where geometric distortion has occurred ... Center of geometric distortion 406 ... Ink splash 407 ... character defect 408 ... pixel 501 ... pair of black defect candidate and white defect candidate 502 ... pair of white defect candidate and black defect candidate

Claims (3)

A printed matter inspection method for inspecting a printed matter by performing image processing after converting the image of the printed matter into digital image data,
[A] A printed matter that has been determined to be good in advance is captured as a reference image, and image data divided into m × n blocks is created.
[B] Taking an image of a printed matter to be inspected as an inspection image, creating image data divided into m × n blocks,
[C] In a region preset for each block, a pattern search is performed with the block of the reference image corresponding to the block to perform a first position correction process,
[D] A difference process and a labeling process are performed between the inspection image subjected to the first position correction process and the reference image, and defect candidates are extracted.
[E] In the vicinity of the defect candidate, if a defect candidate is a white defect candidate, a black defect candidate is searched, and if the defect candidate is a black defect candidate, a white defect candidate is searched within a preset range. And
[F] In the process [e], when there is a defect candidate corresponding to the defect candidate, the number of pixels corresponding to the line width of the defect candidate is calculated, and the divided image region to which the defect candidate belongs is calculated. A position correction amount is set to the number of pixels, a second position correction process is performed on the inspection image, and a difference process and a labeling process are performed between the inspection image subjected to the second position correction process and the reference image. And
[G] In the process [f], if a defect candidate is detected at the same position as the defect candidate, the defect candidate is determined to be a true defect, and if not detected, the defect candidate is determined to be a false detection. And
[H] In the process [e], if there is no defect candidate corresponding to the defect candidate, the defect candidate is determined as a true defect.   The printed matter inspection method according to claim 1, wherein the number of m × n blocks is preset according to the degree of ease of expansion and contraction of the printed matter.   A printed matter inspection apparatus, wherein the printed matter is inspected using the printed matter inspection method according to claim 1.

Images