JP3433331B2 - Image inspection method and apparatus - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image inspection method and apparatus for correcting the displacement between a reference image and an image to be inspected and comparing them for inspection.

[0002]

2. Description of the Related Art Defects in printed matter are caused by defects such as wrinkles, foreign substances, and stains on printing paper, ink stains, hickies, doctor streaks, poor density, poor color tone, etc. that occur during printing. is there. Conventionally, the inspection of the printed matter is mainly a visual inspection performed by a human.For example, in the case of the printed matter discharged from a sheet-fed printing press or a folding unit of the printing press, the operator of the printing press pulls it out at appropriate times. , Visual inspection was performed. If continuous printed matter is ejected from the printing machine and cannot be extracted, the printed matter is irradiated with strobe light that instantaneously emits light in synchronization with the running speed of the printed matter, and the afterimage of the human eye is used to stop the printed matter. Was being visually inspected.

The visual inspection imposes a very heavy work load on the person to be inspected, and since it is performed by a person, it is unavoidable to overlook defects. Therefore, various techniques have been disclosed by making a proposal for automatically inspecting a printed matter. For example, in Japanese Examined Patent Publication No. 1-47823, the image data read from the pattern of the running print at the time when the normal printed matter is printed is stored in the image memory as reference image data, and the reference image data read from the image memory is stored. There is disclosed a technique relating to a printed matter inspection apparatus of a system in which the quality of printing is determined by comparing, in pixel units, with the inspection target image data read from the pattern of the printed matter of the inspection target being printed.

However, when the reference image data and the image data to be inspected are compared on a pixel-by-pixel basis, if the accuracy of defect detection is increased, even if the positional deviation between both data is slight, the positional deviation is erroneously detected as a defect, In order to prevent misregistration of misalignment as a defect, it is necessary to reduce the accuracy of defect detection.If some misalignment cannot be avoided, sufficient inspection performance can be obtained to satisfy the requirements. There wasn't. Therefore, in Japanese Examined Patent Publication No. 1-20477, the inspection target image data or the reference image data is displaced pixel by pixel in the left-right direction with respect to the traveling direction of the printed matter, and the inspection target image data and the reference image data are compared. A technique is disclosed in which the positional deviation is corrected based on the positional deviation amount between the two pieces of data when the data are most matched, and then the inspection target image data and the reference image data are compared.

[0005]

However, in the conventional method for correcting the positional deviation of the image in the automatic inspection technique, the inspection target image data and the reference image data are compared.
There is a problem that an enormous amount of processing time is required because the image data having an extremely large amount of data is processed. SUMMARY OF THE INVENTION It is therefore an object of the present invention to provide an image inspection method and apparatus capable of inspecting an image by performing highly reliable positional deviation correction with a short processing time.

[0006]

The above object can be achieved by the present invention described below. That is, in the image inspection method for inspecting an image to be inspected by comparing with a reference image, which is the first aspect of the present invention, a line segment position which is a position of a pixel row corresponding to a line segment in a direction parallel to a direction in which positional deviation is to be corrected. Selecting a line segment position, and extracting a reference pixel data string on the reference image at the line segment position selected in the line segment position selecting step
Data extraction step and second data extraction step of extracting a plurality of inspection target pixel data strings on the inspection target image from the line segment position and a plurality of positions that are changed in the line segment direction in pixel units from the line segment position And a correlation coefficient calculation process for obtaining a correlation coefficient from the reference pixel data sequence extracted in the first data extraction process and the plurality of inspection target pixel data sequences extracted in the second data extraction process, and the phase thereof. A positional shift amount calculation step of obtaining a positional shift amount from the data of the position of the inspection target pixel data row giving the maximum correlation coefficient among the correlation coefficients obtained in the relation number calculation step and the line segment position data, First Heaven
The position in the up-and-down direction after obtaining the above-mentioned amount of misalignment
Lines in the left and right direction of the pixel data to be detected according to the amount of deviation
Line segment position correction process to correct the segment position and left and right
Based on the corrected line segment position in the left-right direction
To obtain the positional deviation amount, and based on each of the positional deviation amounts.
Then, adjust the displacement in two directions and compare it with the reference image.
An image inspection method including a positional deviation correction comparison process for inspecting an image to be inspected .

In the second aspect of the invention, which is an image inspection apparatus for inspecting an image to be inspected by comparing it with a reference image, a line segment that is a position of a pixel row corresponding to a line segment in a direction parallel to a direction in which misregistration is to be corrected. Line segment position selecting means for selecting a position, first data extracting means for extracting a reference pixel data string on the reference image at the line segment position selected by the line segment position selecting means, the line segment position and its line Second data extraction means for extracting a plurality of inspection target pixel data strings on the inspection target image from a plurality of positions that are position-changed in pixel units from the split position and extracted by the first data extraction means. Correlation coefficient calculating means for obtaining a correlation coefficient from the reference pixel data string and the plurality of inspection object pixel data strings extracted by the second data extracting means, and the correlation coefficient calculating means by the correlation coefficient calculating means. The largest correlation coefficient An inspection object pixel data column position deviation amount calculating means and the position data obtaining position deviation amount from the data of the line positions of the obtaining, not a first said position the vertical direction
The amount of deviation is obtained, and detection is performed according to the amount of displacement in the vertical direction.
A line segment that corrects the horizontal line segment position of output target pixel data
Position correction means and the above-mentioned correction for left and right directions remaining thereafter
Based on the horizontal line segment position
Based on each of the above-mentioned displacement amounts,
Correct this and compare with the reference image to inspect the image to be inspected
An image inspection apparatus including a positional deviation correction comparing unit.

According to a third aspect of the invention , there are a plurality of line segment positions selected by the line segment position selecting means, and the positional shift amount is obtained for each of the line segment positions, and the positional shift amounts are calculated. Second , inspecting the image to be inspected by correcting the positional shift based on the medium-level positional shift amount and comparing it with the reference image
The image inspection apparatus of the invention . According to a fourth aspect of the present invention , in the line segment position selecting means, the reference pixel data string on the reference image at the line segment position selected by the line segment position selecting means has little change in pixel data in a direction perpendicular to the line segment, The image inspection apparatus according to the second or third aspect of the invention , which is a line segment position selecting unit that selects a line segment position in which the pixel data changes largely in the direction parallel to the line segment. According to a fifth aspect of the present invention , in the first data extracting means, the reference pixel data sequence extracted by the first data extracting means is in a direction perpendicular to the pixel data sequence on the reference image at the line segment position and the line segment. The first data extracting means is a pixel data row having pixel data of an average value of pixel data of corresponding pixels of a predetermined number of pixel data rows adjacent thereto, and the second data extracting means is the first data extracting means. 2 The plurality of inspection target pixel data strings extracted by the data extracting means are arranged in the direction perpendicular to the line segment for each of the line segment position and each of the plurality of positions that are changed in the pixel unit from the line segment position in the pixel unit direction. A second data extraction unit that is a plurality of pixel data strings that obtains an addition average value of pixel data for each corresponding pixel of a predetermined number of pixel data strings adjacent to that position for each position, and that is the pixel data string. There, the image inspection device of the second to fourth invention.

[0009]

In the image inspection method and apparatus of the present invention, the reference image and the image to be inspected are corrected by comparing their positional deviations and then compared. The method of correcting the positional deviation is to first select a line segment position that is the position of the pixel column corresponding to the line segment in the direction parallel to the direction to be corrected, and then set the reference pixel data sequence on the reference image at that line segment position. Then, a plurality of inspection target pixel data strings on the inspection target image are extracted from the line segment position and a plurality of positions in which the line segment position is changed pixel by pixel in the line segment direction. Then, a correlation coefficient is obtained from the reference pixel data sequence and the plurality of inspection target pixel data sequences, and the position data of the inspection target pixel data sequence that gives the maximum correlation coefficient among the correlation coefficients and the data The displacement amount is obtained from the data of the line segment position. The amount of misalignment is first measured in the vertical direction.
And the detected pair according to the amount of vertical displacement.
The position of the line segment in the left-right direction of the phantom pixel data is corrected. After that, the positional deviation is corrected based on the positional deviation amounts in the two directions and compared with the reference image to inspect the inspection target image.

According to the image inspection method and apparatus of the present invention,
Processing two-dimensional image data with extremely large data volume
To process one-dimensional image data rather than
In addition, the processing time can be extremely short, and in general
The amount of displacement is large.
Therefore, it is highly reliable.

Further, there are a plurality of the line segment positions selected, the positional shift amounts are obtained for the respective line segment positions, and the positional shift amount of the median value of the positional shift amounts is based on the positional shift amount. By inspecting the image to be inspected by correcting the displacement and comparing it with the reference image, the correct correction amount with high reliability can be obtained even if some of the displacement correction errors occur due to factors such as disturbances. Obtainable.

Further, in the reference pixel data string on the reference image at the line segment position selected in the selection of the line segment position, there is little change in the pixel data in the direction perpendicular to the line segment, and the pixel data is arranged in the direction parallel to the line segment. By selecting the position of the line segment where the change is large, it is less susceptible to the misalignment of the line segment and the vertical direction, and the data amount when calculating the correlation amount is guaranteed,
A correct correction amount with high reliability can be obtained.

The reference pixel data string extracted by the first data extracting means is a pixel data string on the reference image at the line segment position and a corresponding pixel of a predetermined number of pixel data strings parallel to the line segment. A plurality of inspection object pixel data strings extracted by the second data extracting means are pixel data strings in which the arithmetic mean value of the pixel data for each pixel data is used as pixel data. For each of a plurality of positions whose positions have been changed in the direction of the line segment, the arithmetic mean value of the pixel data for each corresponding pixel of the predetermined number of pixel data rows parallel to the position is obtained for each position in the direction perpendicular to the line segment. Since it is a plurality of pixel data strings using it as pixel data, it is possible to reduce the influence of misalignment in the vertical direction with respect to the line segment and obtain a correct correction amount with high reliability.

[0014]

When a target image is, for example, a printed matter, a lateral shift due to meandering or the like and a vertical shift due to tension fluctuation or the like cannot be avoided due to the accuracy of the traveling system of the printed matter in the printing press. Since the collation inspection of these two images having the positional displacement causes the erroneous detection, it is necessary to perform the collation inspection after performing the positional displacement correction. In the conventional misregistration correction, the correlation value of one image as a whole or a partial area and the other image is calculated, and the position having the highest correlation is used as the correction amount. However, these methods have a large amount of processing and have been a problem in terms of processing speed. The present invention is made in order to improve the processing speed in the positional deviation correction performed at the time of image inspection, to significantly reduce the processing amount, and not to impair the reliability.

FIG. 1 is a flow chart showing the flow of data processing in the image inspection method and apparatus of the present invention.
In FIG. 1, first, a line segment position, which is a position of a pixel column corresponding to a line segment in a direction parallel to the direction to be corrected, is selected. The line segment position selected may be one or plural,
This line segment position selection refers to the reference image, and the reference pixel data string on the reference image at the line segment position has little change in pixel data in the direction perpendicular to the line segment and changes in pixel data in the direction parallel to the line segment. Is selected (S1 line segment position selection). This S1 is performed in the line segment position selection process and the line segment position selection means.

Next, the reference pixel data string on the reference image at the line segment position selected in S1 is extracted. At that time, the reference pixel data sequence may be a single pixel data sequence on the reference image at the line segment position, and the reference pixel data sequence is the pixel data sequence on the reference image at the line segment position and the line segment. It may be a pixel data string in which the pixel data is the arithmetic mean value of the pixel data for each corresponding pixel of a predetermined number of pixel data strings adjacent to it in the perpendicular direction. The reference pixel data string is extracted by such a method (S2 first data extraction). This S2 is performed in the first data extraction process and the first data extraction means.

Next, a plurality of inspection target pixel data strings on the inspection target image are extracted from the line segment position selected in S1 and a plurality of positions where the line segment direction is changed in pixel units from the line segment position. It At this time, the plurality of inspection target pixel data rows may be the line segment positions on the inspection target image and a single pixel data row whose position is changed in pixel units in the line segment direction from the line segment positions. Of the inspection target pixel data sequence, the line segment position and a plurality of positions in which the line segment direction is changed in pixel units from the line segment position are adjacent to the line segment in the direction perpendicular to the line segment. It may be a plurality of pixel data strings in which the addition average value of the pixel data for each corresponding pixel of the number of pixel data strings is obtained and used as the pixel data. The inspection target pixel data string is extracted by such a method (S3 second data extraction). This S3
Is performed in the second data extraction process and the second data extraction means.

Next, a correlation coefficient is calculated from the reference pixel data string extracted in S2 and the plurality of inspection object pixel data strings extracted in S3 (S4 correlation coefficient calculation). When there are a plurality of reference pixel data strings extracted in S2, a similar correlation coefficient is calculated for each reference pixel data string. This S4 is performed in the correlation coefficient calculation process and the correlation coefficient calculation means.

Next, when the number of reference pixel data strings extracted in S2 is one, the displacement amount that maximizes the correlation coefficient calculated in S4 is used as the displacement amount. Also S2
If there are multiple reference pixel data strings extracted in step 3, the shift amounts that maximize the correlation coefficient for each reference pixel data sequence are compared, and the position shift amount of the middle value of the position shift amounts is shifted. The calculation of the amount is performed (S5 position shift amount calculation). This S5 is performed in the positional deviation amount calculation process and the positional deviation amount calculation means.

Next, the positional deviation between the reference image and the inspection object image is corrected based on the positional deviation amount obtained in S5, and the inspection object image is inspected by comparing with the reference image (S6).
Misalignment correction comparison). This S6 is performed in the positional deviation correction comparison process and the positional deviation correction comparison means.

FIG. 2 is a pictorial diagram showing the flow of data processing for positional deviation correction in the image inspection method and apparatus of the present invention. In FIG. 2, FIG. 2A shows the reference image, and the line segment represented by the dotted line in the figure is the selected line segment position. The line segment position is the position itself indicated by the line segment in the case of printed matter or photographic film.
In the case of image data as digital data, it means the pixel position at the position indicated by the line segment, that is, the pixel address. Since it is indicated by a line segment, it is the address of the pixel column or pixel data column.

Image data as digital data is
The data length of each pixel is 1 byte (8 bits, 0-25
5), or 6 bits (0 to 63) depending on the required accuracy,
10 bits (0 to 1023) or the like can be selected. In the case of color image data other than black and white, the data length is usually tripled corresponding to the three primary colors. Further, the data may be a colorimetric luminance value, RGB value, XYZ value, density value, reflectance, transmittance and the like. But usually
When digital data is obtained from a printed matter or a photographic film and used, a strict colorimetric value is not necessary, and for example, in printing defect inspection, reproducibility and stability of detection data are often sufficient. Therefore, it is possible to use, as image data, light and shade as data detected according to the characteristics of the detection system or the like, or data related to brightness and luminance values.

In FIG. 2, the reference pixel data string at the line segment position of the reference image is as shown in the graph of FIG. 2 (c). 2B shows the inspection target image, and the inspection target pixel data string at the position of the line segment selected in the reference image is as shown in the graph of FIG. 2D. Along with the inspection target pixel data row at the line segment position selected in this reference image, a plurality of inspection target pixel data rows on the inspection target image are obtained from a plurality of positions where the line segment position is changed in pixel units in the line segment direction. Extract.

Then, the correlation coefficient between the reference pixel data sequence and the plurality of inspection target pixel data sequences is calculated. If the inspection target pixel data string having the largest correlation coefficient is selected and the position of the reference pixel data string is subtracted from the position to obtain the difference, the difference is the misregistration correction amount. The misregistration correction may be performed by using a position obtained by subtracting the misregistration correction amount from the position of the inspection target pixel data string as the position of the inspection object pixel data string after the misregistration correction. As described above, this position has the same meaning as the address of the pixel data or the pixel data string.

In the above description, it has been described that a plurality of inspection object pixel data strings are extracted, but only the inspection object pixel data string at the line segment position selected in the reference image is extracted, and the reference pixel data string and the inspection are extracted. A method of calculating the correlation coefficient of the target pixel data string while shifting one side may be used. In that case, the pixel data protruding by shifting is ignored, and the correlation coefficient is calculated for the pixel data string common to the reference pixel data string and the inspection target pixel data string, and the amount of pixel data to be calculated is different. If the pixel data string is large, the result will not be a big difference. In addition, even if the method includes performing calculation while shifting the reference pixel data string by extracting extra inspection target pixel data strings by the number of pixels to be shifted in the pixel data string including the selected line segment position on the reference image. The amount of pixel data to be calculated is the same as in the case of extracting the plurality of inspection target pixel data strings described above. There is no difference in the results obtained.

In the above description, the case in which the direction to be corrected is one has been described, but in the printed matter or the like being printed, a positional shift occurs in the horizontal direction as well as the vertical direction. When the directions to be corrected are two directions, the amount of positional deviation is obtained by the method described above for one of the two directions, and the remaining 1
It suffices to obtain the positional deviation amount in the same manner for the directions and correct the positional deviation in two directions based on the positional deviation amount. Further, the line segment position selected is not limited to one, but may be plural. It is possible to obtain the positional deviation amount for each of the line segment positions by the above-described method, and correct the positional deviation based on the median positional deviation amount of the positional deviation amounts. If a plurality of line segment positions are provided, the reliability and accuracy are increased as compared with the case where there is one line segment position. Even if the average value is obtained instead of the median value and the fraction with respect to the pixel position is rounded off, the same effect can be obtained. Alternatively, a method of obtaining an average value of the remaining positions of the plurality of line segments other than the maximum and minimum of the line segment positions may be used, and a method of selecting three line segments and obtaining an average value of two close distances May be

Further, the selection of the line segment position has an appropriate position depending on the picture content of the image. By selecting a line segment position in which the reference pixel data string on the reference image at the line segment position has a small change in pixel data in the direction perpendicular to the line segment and a large change in pixel data in the direction parallel to the line segment, Reliability and accuracy will increase. As a method of finding such a position, for example, the edge of the pattern of the image is extracted to avoid a line segment position where an edge substantially parallel to the line segment is near
This can be done by selecting a line segment position where an edge that is substantially orthogonal to the line segment exists.

As the reference pixel data string to be extracted,
It is a pixel data string in which a pixel data string on a reference image at a line segment position and a pixel data of an average value of pixel data of corresponding pixels of a predetermined number of pixel data strings adjacent to the line segment in a direction perpendicular to the line segment are pixel data. In the same manner, as a plurality of inspection target pixel data strings that are similarly extracted, a line segment position and a plurality of positions that are changed in the line segment direction in pixel units from the line segment position It is possible to obtain an arithmetic mean value of pixel data for each pixel corresponding to a predetermined number of pixel data strings adjacent to the position for each position, and make the pixel data a plurality of pixel data strings. By doing so, reliability and accuracy will be increased.

In the conventional misregistration correction, in contrast to the above-mentioned misregistration correction, two-dimensional correlation calculation had to be performed, so that high-speed processing was difficult even if a high-speed processor was used. . However, in the present invention, since the correlation calculation is performed in the one-dimensional array, high speed processing is possible. For example, a DSP (digital
When a signal processor) or the like is used, the correlation calculation of a one-dimensional array can be executed instantaneously.

FIG. 3 is a schematic diagram showing an outline of the apparatus configuration of the present invention. In FIG. 3, 1 is a printing unit, 2 is a printed matter, 3 is a guide roller, 4 is a light source for illumination, 5 is a camera unit for imaging, 6 is a camera control unit for controlling the camera unit, and 7 is a machine of the printing unit. A rotary encoder for detecting a target position (rotational position of the plate cylinder), and 8 is a main body of a printed matter inspection apparatus for processing image data obtained by imaging by a camera unit. The printing paper passes through the printing unit 1, where printing is performed to form the printed matter 1, which is guided by the guide roller 3 and reaches a position illuminated by the light source 4. There is usually a guide roller 3 at that position, which defines the position of the printed matter 2. The printed matter 2 is imaged by the camera unit 5 at that position. The camera unit 5 is a linear sensor camera that captures an image by one-dimensional scanning in the example of FIG.

A rotary encoder 7 is installed in the plate cylinder of the printing unit 1 or in a portion that rotates in synchronization with the plate cylinder, and outputs a pulse signal in synchronization with the printing cycle.
This pulse signal is input to the camera control unit 6, and the camera control unit 6 outputs an imaging start signal to the camera unit 5 every time the printed matter 2 travels a predetermined amount in synchronization with the printing cycle, and the camera unit 5 captures an image. Data, that is, analog data is input, A / D (analog to digital) conversion is performed to convert it into digital data, and the data is transferred to the printed matter inspection apparatus main body 8. In this way, two-dimensional scanning is performed at a timing synchronized with the printing cycle, the digital image data is read into the printed matter inspection apparatus main body 8, and since the image is captured in synchronization with the printing cycle in this manner, If there is no disturbance factor or error factor, the imaging data read in each printing cycle will always be the imaging data at the same print pattern position if the imaging data is at the same cycle timing.

The printed matter inspection apparatus main body 8 to which the digital image signal is input stores the digital image data of the printed matter 1 when a non-defective product is printed in the memory as a reference image. Then, it is determined whether or not a defect of the printed material 1 has occurred by using the digital image data input thereafter as the inspection target image data.

FIG. 4 is a block diagram showing the structure of the main body of the printed matter inspection apparatus. In FIG. 4, 8 is a printed matter inspection apparatus main body, 5 is a camera unit, 6 is a camera control unit, 9 is a reference image data memory, 10 is an inspection target image data memory, 11 is a reference input command input unit, and 12 is a displacement detection. Reference numeral 13 is a positional deviation correction comparison unit, and 14 is a parameter input unit. Each of the reference image data memory 9 and the inspection target image data memory 10 has a storage capacity capable of storing at least one unit of printed matter (a printing screen for one cycle performed by the printing unit). Further, the reference input command input unit 11 is a command for taking a reference image to start the image inspection and inputting data to the reference image data memory 9, and then taking an inspection target image to be stored in the inspection target image data memory 10. Switch the command to input.

When the reference image data is input by the reference input command input unit 11, the selector shown in FIG. 2 connects the image data to the reference image data memory 9, and at this time, the positional deviation detection unit 12, And the positional deviation correction comparison unit 13 does not operate. On the other hand, when the input of the inspection target image data is commanded by the reference input command input unit 11, the selector makes an input connection of the imaging data to the inspection target image data memory 10. Then, the positional deviation detecting unit 12 detects the positional deviation between the inspection target image and the reference image, and sends the resulting positional deviation correction amount to the positional deviation correction comparing unit 13. Then, the misregistration correction comparison unit 13 that has received the misregistration correction amount performs a collation comparison inspection of the inspection target image data and the reference image data in consideration of the misregistration correction amount. In addition, the parameter input unit 14
Is the positional deviation detection unit 12 and the positional deviation correction comparison unit 1
Input for setting the parameters used in 3. As the parameter, for example, a threshold value is input when the misregistration correction comparison unit 13 compares the inspection target image data with the reference image data to determine whether the image is good or bad.

FIG. 5 is a block section showing the structure of the position shift detecting section. In FIG. 5, reference numeral 12 denotes a positional deviation detection unit,
Reference numeral 9 is a reference image data memory, 10 is an inspection target image data memory, 11 is a reference input command input unit, 13 is a positional deviation correction comparing unit, 15 is a line segment position selecting unit, 16 is a first data extracting unit, and 17 is a first data extracting unit. 2 is a data extraction unit, 18 is a first correlation line segment memory, 19 is a second correlation line segment memory, 20 is a correlation coefficient calculation unit, and 21 is a median value calculation unit. Further, FIG. 6 is a flow chart showing the flow of data processing in the positional deviation detecting unit 12.

In FIGS. 5 and 6, the line segment position selection unit 15 first refers to the reference image data in the reference image data memory 9 and selects five line segment positions in the vertical direction (S1). First
The data extraction unit 16 of (1) inputs the line segment position and extracts the reference pixel data string of each line segment position in the vertical direction from the reference image data memory 9 (S2). In addition, the second data extraction unit 1
7 inputs the line segment position and extracts a plurality of inspection target pixel data strings at each line segment position in the vertical direction from the inspection target image data memory 10 (S3).

One of the reference pixel data strings extracted by the first data extraction unit 16 is stored in the first correlation memory 18. One of the inspection target pixel data strings extracted by the second data extracting unit 17 is the second correlation memory 1
9 is stored. The correlation coefficient calculation unit 20 calculates a correlation coefficient between the reference pixel data sequence stored in the first correlation memory 18 and the second correlation memory 19 and the inspection target pixel data sequence. Such calculation of the correlation coefficient is repeated a plurality of times for a plurality of inspection target pixel data rows corresponding to each reference pixel data row to obtain a plurality of correlation coefficients (S4).

The median value calculation unit 21 includes a correlation coefficient calculation unit 20.
Among the plurality of correlation coefficients calculated in step 1, the inspection target pixel data string that gives the maximum correlation coefficient for each line segment position is obtained,
The amount of displacement is calculated (S5). Further, the median value calculation unit 21 obtains the median value from the positional deviation amount for each line segment position obtained by the correlation coefficient calculation unit 20 and sets it as the positional deviation amount in the vertical direction. The positional deviation amount in the vertical direction is output to the positional deviation correction comparing unit 13 (S6). The above S1 to S6
The step of obtaining the amount of positional deviation in the vertical direction described above can be performed in the same manner in the horizontal direction. Therefore, the description of steps S7 to S12 for obtaining the amount of positional deviation in the left-right direction will be omitted.

As described above, in this example, after the positional deviation in the vertical direction is corrected, the lateral positional deviation is corrected.
When inspecting printed matter, the behavioral state of the printed matter due to running is more than the positional displacement in the left-right direction due to meandering. The displacement of is usually larger. Therefore, as in this example, it is possible to perform the positional deviation correction with higher reliability by performing the positional deviation correction in the vertical direction first.

In this example, the number of line segment positions is 5 in both the vertical direction and the horizontal direction. Since the median value of these five misregistration correction values is used as the final misregistration correction value, a correct correction value can be obtained even if an error occurs in up to two of the five results. In this way, by using a plurality of line segment positions, it is possible to improve the reliability of positional deviation correction. The selection of such a line segment position by the line segment position selection unit 15 is performed in the vertical direction line segment in which there is little data change when the position shifts in the left and right directions from the reference image data, and in the left and right direction where the data change does not change when the position shifts in the up and down direction. Select a line segment.

In the first data extraction unit 16 and the second data extraction unit 17, the reference pixel data sequence and the detection target pixel data of the line segment position selected by the line segment position selection unit 15 are used as a one-dimensional array to make the first correlation. It is stored in the line segment memory 16 and the second correlation line segment memory 17. However, when the line segment position in the left-right direction is selected by the line segment position selection unit 15, as shown in FIG. 7, since the positional deviation correction amount in the vertical direction is known, detection is performed according to the positional deviation correction amount. The position of the line segment in the left-right direction of the target pixel data is corrected. FIG. 7 is a diagram showing that the position of the line segment in the left-right direction of the detection target pixel data is corrected according to the amount of misalignment correction in the vertical direction. The first correlation line segment memory 16 and the second correlation line segment memory 17 are memories for storing a reference pixel data sequence and an inspection target pixel data sequence corresponding to the number of pixels included in the line segment position. It has a capacity of several (five in this example).

The correlation coefficient calculation unit 20 obtains the correlation coefficients of the reference pixel data sequence and the plurality of inspection target pixel data sequences, respectively, and obtains the shift amount that maximizes the correlation coefficient. At that time, the DC data components of the reference pixel data sequence and the inspection target pixel data sequence are removed in order to increase the change in the correlation coefficient according to the shift. That is, the average value of each of the reference pixel data row and the inspection pixel data row is subtracted from each pixel data. When this is expressed by a mathematical formula, it becomes as shown in Formula 1.

F ′ (i) = f (i) −f ^ g '(i) = g (i)-g ^ However, f '(i); new reference pixel data string g '(i); New inspection target pixel data string i; position (address) f (i); reference pixel data string g (i); inspection target pixel data string f ^; average value of f (i) g ^; average value of g (i)

The correlation coefficient calculated by the correlation coefficient calculation unit 20 is calculated by the following equation 2.

[Equation 2] However, Rj; correlation coefficient j; position of inspection target pixel data string n; number of pixels in pixel data string Mathematically, the correlation coefficient is f
This is a product of the standard deviation of (i) and g (i) and the product of n, and dividing the Rj by the product. However, a mathematically exact correlation coefficient is not necessary here, and only the relative magnitude relationship is important, so the value obtained by the simplified Equation 2 is used as the correlation coefficient. Further, in the correlation coefficient calculation unit 20, R
The value of the position j of the inspection target pixel data string where j is the maximum is calculated, and this is used as the positional deviation detection amount at each line segment position.

The median value calculation unit 21 calculates the median value of the amount of displacement detection at the position of five line segments. For example,
It is assumed that the positional deviation detection amount at each line segment position obtained by the correlation coefficient calculation unit 20 is as follows. {3,4,3,1,5} When rearranged in descending order of value, the result is as follows. {5,4,3,3,1} From this, a median value of 3 is obtained and used as the positional deviation detection amount used for actual positional deviation correction. In the correlation coefficient calculation unit of this example, the correlation coefficient is used to obtain the degree of coincidence between the two line segments, but the degree of coincidence is calculated by summing the absolute values of the differences between the two data strings. You may judge.

FIG. 8 is a flow chart showing the flow of data processing when the line segment position selection unit 15 selects a line segment in the vertical direction. As for the line segment position in the vertical direction, the influence of the left and right misalignment is small, that is, the place of the reference image with few horizontal edges is preferentially selected. In FIG. 8, first, differential image data in the horizontal direction is calculated from the reference image data (S101). The difference image data is calculated by assuming the maximum width of left / right positional deviation. That is, the pixel value Pi of the difference image is represented by the following Expression 3.

## EQU00003 ## Pi = max | Pi + i'-Pi | However, the above function max is an element when i'is changed in the range of -k≤i'≤ + k, where k is the maximum width of the lateral displacement. It shows that it takes the maximum value in the group.

Next, the calculated horizontal difference image is binarized. The binarization is performed by using a predetermined threshold value. For example, a calculated value such as an average value of all pixels may be used as the threshold value (S102). Next, each pixel of the horizontal difference binary image is added in the vertical direction, and the histogram is calculated. This histogram represents the number of pixels having horizontal edges in each vertical direction (S103). next,
A vertical difference image is calculated from the reference image. This vertical difference image is used to guarantee the number of edges in the vertical direction to some extent, and is calculated by the difference between the data of the target pixel and the data of the next pixel (S104). The vertical difference image is binarized similarly to the horizontal difference image (S105), and is further added in the vertical direction to calculate a histogram. This histogram represents the number of pixels having vertical edges in each vertical direction (S106).

Next, in the horizontal difference histogram, the horizontal position having the minimum value is selected (S107). Then, it is determined whether the value of the vertical difference histogram corresponding to the horizontal position falls within a certain range (S10).
8) If there is, the horizontal position is set as the position of the vertical correlation line segment (S109). If the value of the vertical difference histogram is not within the certain range (S108),
Select the next minimum value of the horizontal difference histogram (S1
07), the same determination is made. These processes are repeated until five vertical direction correlation line segments are selected (S110). In addition, make sure that the selected vertical line segments are not next to each other.
It is preferable to invalidate the left and right several lines that have been selected once so that they will not be selected in the next process. The reason is that if the line segment positions are adjacent to each other, one printed matter defect may occur on several line segments at a time, and the pixel data strings of all of them may change. Is.

FIG. 9 is a flow chart showing the flow of data processing when the line segment position selection unit 15 selects a line segment position in the horizontal direction. Basically, there is no difference from the case where the line segment position in the vertical direction in FIG. 8 is selected. In the description of FIG. 8, if the vertical and horizontal directions are read, the left and right are vertical and the horizontal is vertical, and the horizontal is vertical. The description is omitted because it is similar.
In FIG. 9, the pixel value Pj of the difference image is expressed by the following Expression 4 instead of Expression 3.

## EQU00004 ## Pj = max | Pj + j'-Pj | However, the above function max is an element when j'is changed in the range of -k ≤ j'≤ + k, where k is the maximum width of the lateral displacement. It shows that it takes the maximum value in the group.

FIG. 10 is a block diagram showing the structure of the positional deviation correction comparison unit. In FIG. 10, 13 is a positional deviation correction comparison section, 6 is a camera control section, 9 is a reference image data memory, 10 is an inspection target image data memory, 12 is a positional deviation detection section, 22 is an expansion processing circuit, and 23 is degeneration processing. circuit,
24 is an upper threshold setting circuit, 25 is a lower threshold setting circuit, 2
6 is an upper threshold image buffer, 27 is a lower threshold image buffer, 28 is an image shift circuit, 29 is an upper threshold comparison circuit,
Reference numeral 30 is a lower threshold comparison circuit. In the above configuration,
The image shifting circuit 28 includes the inspection target image data memory 10
The image is shifted according to the positional deviation correction amount obtained by the positional deviation detection unit.

The degeneracy processing circuit 23 performs a process of narrowing the area corresponding to the bright area of the image data written in the reference image data memory 9. vice versa,
The expansion processing circuit 22 performs processing for thickening a bright area. As an example of the actual processing, the degeneration processing circuit 23 is expressed by the following Expression 5, and the expansion processing circuit 22 and the following Expressions 6.

F (x, y) = min {f (x-1, y-
1), f (x, y-1), f (x + 1, y-1), f
(X-1, y), f (x, y), f (x + 1, y), f
(X-1, y + 1), f (x, y + 1), f (x + 1,
y + 1)} However, the function min {} indicates that the minimum value of the listed variables is taken.

## EQU6 ## f (x, y) = max {f (x-1, y-
1), f (x, y-1), f (x + 1, y-1), f
(X-1, y), f (x, y), f (x + 1, y), f
(X-1, y + 1), f (x, y + 1), f (x + 1,
y + 1)} However, the function max {} indicates that the maximum value among the listed variables is taken.

Expression 5 represents that the darkest brightness value is newly set as the brightness of the target pixel from the brightness of the target pixel and the pixels in the vicinity of the target pixel. In addition, Equation 6 is the target pixel and its 8
It represents that the brightest brightness value is newly set as the brightness of the target pixel from the brightness of the neighboring pixels. The lower limit threshold value setting circuit 25 has a lower limit value of the brightness variation set in advance by the parameter input unit 14, and the lower limit value is subtracted from the output of the degeneration processing circuit 23.
Written in. Similarly, in the upper limit threshold setting circuit 24, the upper limit value of the brightness variation is set in advance from the parameter input unit 14, and the upper limit value is added to the output of the expansion processing circuit 22 and written in the upper limit threshold image buffer 26. .

In the lower threshold comparison circuit 30, the luminance value of the lower threshold image buffer 27 and the inspection target image data memory 10 are compared.
Are compared for each corresponding pixel, and when the brightness value of the inspection target image data memory 10 is lower than the brightness value of the lower limit threshold image buffer 27, it is determined that a defect has occurred.
The upper threshold comparison circuit 29 compares the luminance value of the upper threshold image buffer 26 and the luminance value of the inspection target image data memory 10 for each corresponding pixel, and the luminance value of the inspection target image data memory is compared with the upper threshold image buffer. If the brightness value is higher than 26, it is determined that a defect has occurred.

Since the positional deviation correction processing of the present invention is correction for each pixel, the positional deviation of less than one pixel cannot be corrected even if the correction processing is performed. Therefore, the misregistration correction comparison unit 13 is provided with the degeneration processing circuit 23 and the expansion processing circuit 22 to allow erroneous detection at an edge portion in the image where the luminance change is large. Further, the positional deviation correction comparison unit 1 of the present invention
The configuration of 3 is not limited to this embodiment, and the brightness of each pixel of the reference image memory and the brightness of each pixel of the image to be inspected may be simply compared according to the amount of positional deviation.

[0054]

As described above, according to the image inspection method and apparatus of the present invention, first, the reference pixel data string on the reference image in the direction parallel to the direction to be corrected is extracted and the position thereof is further corresponded. To extract a plurality of inspection target pixel data strings.
Then, the correlation coefficient is obtained from the data strings, and the positional deviation amount is obtained from the data of the position of the inspection target pixel data string that gives the maximum correlation coefficient. The amount of displacement is first in the vertical direction.
According to the amount of displacement in the vertical direction,
The position of the line segment in the horizontal direction of the detection target pixel data is corrected.
After that, the positional deviation is made based on the positional deviation amounts in those two directions.
The inspection target image by correcting the
It Therefore, two-dimensional image data that has an extremely large amount of data is
Processing one-dimensional image data instead of processing data
Therefore, the processing time can be extremely short, and
Generally, the amount of misalignment is large.
High reliability because it is done. Further, the reference pixel data string on the reference image can be selected so that the pixel data change is small in the vertical direction and the pixel data change is large in the parallel direction. Hateful. Furthermore, if the reference pixel data string and the inspection target pixel data string use the arithmetic mean value of the pixel data for each corresponding pixel of a predetermined number of pixel data strings adjacent to them, the effect of misalignment in the line segment and vertical direction Is more difficult to receive. Therefore, a correct correction amount with high reliability can be obtained, a data amount for calculating the correlation amount is guaranteed, and a correct correction amount with high reliability can be obtained. That is,
According to the present invention, it is possible to perform a high-speed misregistration correction process for inspection, and since the misregistration correction process is also extremely reliable, high-speed inspection is possible, and inspection accuracy and reliability are improved. effect called excellent can be obtained.

[Brief description of drawings]

FIG. 1 is a flowchart showing a flow of data processing in an image inspection method and apparatus of the present invention.

FIG. 2 is a pictorial diagram showing the flow of data processing for positional deviation correction in the image inspection method and apparatus of the present invention.

FIG. 3 is a schematic diagram showing an outline of a device configuration of the present invention.

FIG. 4 is a block diagram showing a configuration of a printed matter inspection apparatus main body.

FIG. 5 is a flowchart showing a flow of data processing in a position shift detecting unit.

FIG. 7 is a diagram showing that the position of a line segment in the left-right direction of pixel data to be detected is corrected according to a position shift correction amount.

8 is a flowchart showing a flow of data processing when a line segment in the vertical direction is selected by the line segment position selection unit 15. FIG.

FIG. 9 is a flowchart showing a flow of data processing when a line segment position in the vertical direction is selected by the line segment position selection unit 15.

FIG. 10 is a block diagram showing a configuration of a positional deviation correction comparison unit.

[Explanation of symbols]

1 printing unit 2 printed matter 3 guide rollers 4 light sources 5 camera unit 6 Camera control unit 7 rotary encoder 8 Printed matter inspection device body 9 Reference image data memory 10 Image data memory for inspection 11 Standard input Command input section 12 Positional deviation detector 13 Misregistration correction comparison unit 14 Parameter input section 15 line segment position selector 16 First data extraction unit 17 Second data extraction unit 18 First correlation line segment memory 19 Second correlation line segment memory 20 Correlation coefficient calculator 21 Median value calculator 22 expansion processing circuit 23 Degeneration processing circuit 24 upper limit threshold setting circuit 25 lower limit threshold setting circuit 26 upper threshold image buffer 27 Lower threshold image buffer 28 image shift circuit 29 upper threshold comparison circuit 30 lower threshold comparison circuit

─────────────────────────────────────────────────── ─── Continuation of the front page (56) Reference JP-A-59-99592 (JP, A) JP-A-2-217968 (JP, A) JP-A-3-257304 (JP, A) JP-A-3- 41577 (JP, A) JP 3-290786 (JP, A) JP 59-40241 (JP, A) (58) Fields investigated (Int.Cl. ⁷ , DB name) G06T 7/00-7 / 60 B41F 33/14 G06T 1/00 H04N 1/40-1/409 H04N 1/00

Claims (5)

(57) [Claims] 1. An image inspecting method for inspecting an image to be inspected by comparing with a reference image, wherein a line segment position which is a position of a pixel row corresponding to a line segment in a direction parallel to a direction in which positional deviation is to be corrected is selected. A line segment position selection process, a first data extraction process of extracting a reference pixel data string on the reference image at the line segment position selected in the line segment position selection process, the line segment position, and pixels from the line segment position A second data extracting step of extracting a plurality of inspection object pixel data strings on an inspection object image from a plurality of positions whose positions are changed in units of line segments, and a reference pixel data string extracted in the first data extracting step Of the correlation coefficient calculation process for obtaining a correlation coefficient from the plurality of inspection target pixel data strings extracted in the second data extraction process and the correlation coefficient obtained in the correlation coefficient calculation process. Test pair that gives the correlation coefficient Obtaining a positional deviation amount calculating process of obtaining data and positional deviation amount from the data of the line positions of the pixel data string, first the vertical direction the positional deviation amount, the vertical
To the left of the pixel data to be detected according to the amount of positional deviation in the direction
A line segment position correction process for correcting the position of the line segment in the right direction, and the corrected line in the left and right direction for the remaining left and right direction
The position deviation amount is obtained based on the minute position, and
Corrects misalignment in two directions based on the amount of misalignment
Positional deviation correction ratio to inspect the image to be inspected by comparing with the image
An image inspection method comprising: a comparison process . 2. An image inspection apparatus for inspecting an image to be inspected by comparing with a reference image, and selects a line segment position which is a position of a pixel row corresponding to a line segment in a direction parallel to a direction in which positional deviation is to be corrected. Line segment position selecting means, first data extracting means for extracting a reference pixel data string on the reference image at the line segment position selected by the line segment position selecting means, the line segment position and pixels from the line segment position Second data extracting means for extracting a plurality of inspection object pixel data strings on the inspection object image from a plurality of positions whose positions have been changed in the direction of the line segment, and reference pixel data strings extracted by the first data extracting means And a correlation coefficient calculating means for obtaining a correlation coefficient from the plurality of inspection object pixel data strings extracted by the second data extracting means, and the largest of the correlation coefficients obtained by the correlation coefficient calculating means. Test pair that gives the correlation coefficient Obtaining a positional deviation amount calculating means for obtaining data and positional deviation amount from the data of the line positions of the pixel data string, first the vertical direction the positional deviation amount, the vertical
To the left of the pixel data to be detected according to the amount of positional deviation in the direction
Line segment position correcting means for correcting the line segment position in the right direction, and the left and right line that has been corrected in the left and right directions that remain thereafter.
The position deviation amount is obtained based on the minute position, and
Corrects misalignment in two directions based on the amount of misalignment
Positional deviation correction ratio to inspect the image to be inspected by comparing with the image
An image inspection apparatus comprising: a comparison unit . 3. The line segment positions selected by the line segment position selecting means are plural, and the position shift amount is obtained for each of the line segment positions, and the middle position of the position shift amounts is obtained. claims, characterized in that to correct the positional deviation based on the amount of positional deviation value is compared with a reference image to inspect the inspection object image
Item 2. The image inspection apparatus according to item 2 . Wherein said line segment position selection means, the reference pixel data string of the reference image of the selected line segments located by the segment position selection means in line with the vertical direction much variation in the pixel data,
And a line segment position selecting means for selecting a line segment position in which a change in pixel data is large in a direction parallel to the line segment.
The image inspection apparatus according to claim 2 or 3 . Wherein said first data extracting means, the reference pixel data string extracted by the first data extracting means, the pixel data strings on the reference image of the line position and, therewith the line perpendicular direction The first data extracting means is a pixel data row having pixel data which is an addition average value of pixel data of corresponding pixels of a predetermined number of adjacent pixel data rows, and the second data extracting means is the second data extracting means. A plurality of inspection target pixel data strings extracted by the data extraction means are arranged in the direction perpendicular to the line segment for each of the line segment position and each of a plurality of positions changed in the line segment direction in pixel units from the line segment position. A second data extraction unit that is a plurality of pixel data strings having pixel data as pixel data, which is obtained by calculating an addition average value of pixel data of corresponding pixels of a predetermined number of pixel data strings adjacent to the position. The image inspection apparatus according to any one of claims 2 to 4, wherein: