JP2006047077A - Method and device for detecting line defects of screen - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method and a device for detecting the line defects of a screen with high reliability. <P>SOLUTION: Above problems are solved by a detection method of the line defect of a screen, including a sampling region demarcation process for selecting a target pixel from inputted image information on the screen and demarcating a sampling region so that the target pixel is included; a feature pixel extraction process for extracting feature pixels so that the luminance is maximized or minimized for each row of pixels from pixel data at the sampling region; a feature value setting process for setting a value, where the aimed pixel is weighted according to the distance between the predicted position of line defects determined from the position information of the feature pixel and the aimed pixel; and a discrimination process for repeatedly performing processes from the sampling region demarcation process to the feature value setting process for each aimed pixel, for setting a feature value for the entire pixels in the entire region of the screen and discriminating the line defects on the screen, based on the feature value. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a screen line defect detection method and detection apparatus, and more particularly to a highly reliable screen line defect detection method and detection apparatus.

  In recent years, there has been a demand for higher functionality and higher image quality of display devices such as liquid crystal panels and projectors, and development of an automatic inspection method for luminance unevenness that can detect luminance unevenness with high sensitivity and an automatic inspection apparatus that can execute it. Is desired.

  As part of this, an attempt has been made to automatically detect line defects that occur continuously in the vertical or horizontal direction of the screen to be inspected.

  For example, Patent Document 1 discloses an image for extracting a line defect by accumulating pixel values in a vertical direction (vertical direction), a horizontal direction (horizontal direction), or an oblique direction around a target pixel to extract a maximum value. A method for detecting line defects by generating and comparing with a predetermined threshold is disclosed.

Patent Document 2 discloses a method for further improving the detection ability of line defects by correcting the angle of the captured image and distortion of the optical system and performing shading correction.
Japanese Patent Laid-Open No. 10-240933 JP 2003-168103 A

  However, in the above-described conventional method, there is a defect even in a place where the line defect originally does not exist due to an extremely high or extremely low luminance point defect or noise that cannot be removed by the conventional preprocessing. In some cases, the reliability of the detection results was not sufficient.

  Accordingly, an object of the present invention is to provide a highly reliable method and apparatus for detecting line defects on a screen.

  In order to solve the above problem, the present invention selects a pixel of interest from input image information of a screen, defines a sampling region so as to include the pixel of interest, and pixel data of the sampling region A feature pixel extraction process for extracting a feature pixel having a maximum or minimum luminance for each pixel column, and a distance between a predicted position of a line defect obtained from position information of the feature pixel and the target pixel in the target pixel. A feature value setting process for setting a weighted value in accordance with and a process from the sampling area defining process to the feature value setting process are repeated for each pixel of interest, so that all pixels in the entire area of the screen are processed. Providing a method for detecting a line defect on a screen, including a determination process for setting a feature value and determining a line defect on the screen based on the feature value Is shall.

  According to this, feature pixels that are candidates for line defects (also referred to as linear defects) with the maximum or minimum luminance are extracted for each pixel column, and feature values based on distance information with respect to the feature pixels are obtained for all pixels. By adding, the influence of point defects other than line defects (also called point defects) and noise can be reduced, so that line defects can be detected with high sensitivity, and the reliability of the detection results obtained can be improved. Is also expensive.

  Here, for example, when detecting a so-called white line defect (hereinafter also referred to as a white line defect), it is only necessary to extract the one having the maximum luminance in the feature pixel extraction process, and a so-called black line defect (hereinafter referred to as a black line defect). In the case of detecting a line defect), it is sufficient to extract the one having the minimum luminance in the feature pixel extraction process.

  For example, when detecting a line defect extending in the X-axis direction (hereinafter, also referred to as an X-ray defect), a pixel column perpendicular to the X-axis (a pixel column parallel to the Y-axis) in the feature pixel extraction process ), A feature pixel having a maximum or minimum luminance is extracted. For example, when detecting a line defect extending in the Y-axis direction (hereinafter also referred to as a Y-line defect), a pixel column perpendicular to the Y-axis (a pixel column parallel to the X-axis) in the feature pixel extraction process ), A feature pixel having a maximum or minimum luminance is extracted.

  Specifically, as the feature value setting process, for example, a process of creating a virtual line from the feature pixel, a process of calculating a distance between the virtual line and the target pixel, and weighting according to the distance are performed. A process including a process of setting the performed value to the pixel of interest. According to this, by creating a virtual line from a feature pixel, it is possible to draw a virtual line at a position where there is a high probability of a line defect, and a feature value with high accuracy can be set by the pixel of interest.

  Here, as a method of creating the virtual line, specifically, a method of approximating values aligned in the vertical direction with respect to the pixel column by a least square method, a method of calculating a first moment, or the like can be used.

  The feature value setting process includes a process of calculating a distance between a reference line that includes the target pixel and is perpendicular to the pixel column, and each of the feature pixels, and a value that is weighted according to the distance. A process including the process of setting the pixel of interest. According to this, since the process of creating a virtual line becomes unnecessary, it is possible to set a feature value with good accuracy by a relatively simple process for the pixel of interest.

  Preferably, in the determination process, the process from the sampling area defining process to the feature value setting process is repeated for each pixel of interest, thereby creating a feature value map in which feature values are set for all pixels in the entire area. And a step of determining a line defect on the screen based on the feature value map. Thus, by creating the feature value map, it becomes easier to determine the line defect.

  Preferably, the step of determining the line defect on the screen detects the line defect by integrating the feature values based on the feature value map and comparing a maximum value of the integrated values with a predetermined threshold value. It is a process. This makes it possible to make a simple determination.

  Preferably, the process of determining a line defect on the screen includes a dividing process of dividing the feature value map in a uniaxial direction, and integrating the feature values for each of the divided areas to obtain a maximum integrated value. Line defect detection for detecting a line defect by comparing the maximum value calculation process to be obtained and the maximum value of the integrated values of the respective divided areas, and comparing the maximum value among the maximum values with a predetermined threshold value. Process. Thereby, it is possible to detect a line defect having a short line length.

  Preferably, the step of determining a line defect on the screen defines a first integration range in which a predetermined integration range is defined in the feature value map, and the feature value is integrated within the integration range to obtain a maximum integrated value. A value calculating process, a second maximum value calculating process for setting the next integrated range so as to partially overlap the previous integrated range, and obtaining a maximum value of the integrated value of the feature value for the next integrated range; By repeating the second maximum value calculation process, integration is performed over the entire range in the feature value map, the maximum value of the integrated values of each integrated range is compared, and the largest value among the maximum values is determined. And a line defect detection process of detecting a line defect by comparing with a predetermined threshold value. When there is a short line defect at the division position, it is assumed that the line defect is further shortened and cannot be detected by dividing the image. However, such inconveniences can be avoided by integrating the divided images partially overlapping in this way.

  Preferably, before the sampling area defining step, an image compression step of compressing the image information in a uniaxial direction to obtain a compressed image is included. According to this, detection is possible even when the line width of the line defect is large.

  Preferably, the image compression process is a process of compressing the image information at at least two compression rates to obtain at least two compressed images. According to this, even when a line defect having a plurality of line widths exists, it can be reliably detected.

  Preferably, in the sampling area defining step, when the sampling area is determined so that the pixel of interest is normally located in the approximate center, and the pixel of interest is determined to be approximately in the center of the sampling area, the sampling area is defined as The position of the sampling area is determined so that the end of the sampling area coincides with the end of the entire area so that the sampling area is accommodated in the entire area only when protruding from the entire area to be inspected. According to this, even when the sampling region protrudes from the entire region, it is possible to provide highly accurate information to the pixel of interest in a later process by a relatively simple process.

  Preferably, the method further includes an extraction step of extracting a region to be inspected from the input screen image information before the sampling region defining step. For example, information other than the screen may be included in the image information, for example, when a projected video is shot, and there is a possibility that the detection sensitivity may be lowered. According to this, such inconvenience is avoided. It becomes possible.

  According to another aspect of the present invention, a pixel of interest is selected from image information of an input screen, sampling region demarcating means for demarcating a sampling region so as to include the pixel of interest, and each pixel column from pixel data of the sampling region A feature pixel extracting means for extracting a feature pixel having a maximum or minimum luminance every time, and weighting according to a distance between an expected position of a line defect obtained from position information of the feature pixel and the target pixel Characteristic value setting means for setting the value obtained by performing the determination, including a determination means for setting a characteristic value for each pixel of interest in the entire area of the screen and determining a line defect on the screen based on the characteristic value This is a screen line defect detection device.

  According to this, feature pixels that are candidates for line defects (also referred to as linear defects) with the maximum or minimum luminance are extracted for each pixel column, and feature values based on distance information with respect to the feature pixels are obtained for all pixels. Since the influence of point defects (also referred to as point defects) and noise other than line defects is reduced, the line defects can be detected with high sensitivity.

  Specifically, the feature value setting unit includes a virtual line creation unit that creates a virtual line from the feature pixel, a distance calculation unit that calculates a distance between the virtual line and the pixel of interest, and the pixel of interest. Setting means for setting a weighted value according to the distance. According to this, by creating a virtual line from a feature pixel, it is possible to draw a virtual line at a position where there is a high probability of a line defect, and a feature value with high accuracy can be set by the pixel of interest.

  The feature value setting means includes a reference line that includes the pixel of interest and is perpendicular to the pixel column, a distance calculation means that calculates a distance between each feature pixel, and a value that is weighted according to the distance. And setting means for setting the pixel of interest. According to this, since a means for creating a virtual line is not required, it is possible to set a feature value with good accuracy by a relatively simple means for the pixel of interest.

  Preferably, the determination unit sets a feature value for each pixel of interest in the entire region, and then creates a feature value map based on the feature value. This makes it easier to determine line defects.

The present invention will be described below with reference to the drawings.
FIG. 1A is a configuration diagram illustrating an example of a screen line defect detection apparatus according to the present invention. In the present embodiment, a case where the screen 110 in which the liquid crystal light valve is projected onto the screen 120 by the projector 100 is inspected will be described.

  The line defect detection apparatus according to the present embodiment includes an imaging unit 200 for imaging a screen 110 projected on a screen 120 and image information for detecting a line defect from image information (image data) obtained from the imaging unit 200. The processing unit 300 and an output unit 400 for outputting detection information obtained by the image information processing unit 300 are mainly configured.

  The image information processing unit 300 includes an input unit 310, an inspection target region extraction unit 320, a background image difference processing unit 330, a smoothing processing unit 340, an image duplicating unit 350, and a line defect detecting unit 360. It is configured. The image information processing unit 300 is usually configured by a central processing unit (CPU).

  Image information obtained by the imaging unit 200 is input to the input unit 310 as a digital signal, and this image information is stored in a storage unit (not shown). The inspection target region extraction unit 320 extracts a region to be inspected (inspection target region) from the image information of the acquired captured image. That is, when the screen 110 projected on the screen 120 is imaged, a part of the screen 120 on which no image is projected may be captured in addition to the projected screen 110. In such a case, for example, only the screen 110 portion to be inspected is extracted by detecting the coordinates of the four corners of the image to be inspected based on the captured image information. At this time, if the captured screen 110 is distorted or tilted, a process of reconstructing the image area into a rectangular shape may be performed by applying geometric deformation to the inspection target area. . The background image difference processing unit 330 samples representative values from several locations in the inspection target image area, creates a background image by interpolating between the respective sampling values by linear interpolation or spline interpolation, and the like. Shading correction such as subtracting or dividing from. The smoothing processing unit 340 uses a spatial filter such as an averaging filter or a median filter, and removes noise components caused by characteristics of the imaging unit 200 and the like. The image duplicating unit 350 duplicates an image to be processed according to the number of types of processing performed by the line defect detecting unit 360 and stores the image in a storage unit (not shown). As shown in FIG. 1B, the line defect detection unit 360 is mainly composed of a sampling area demarcating unit 362, a feature pixel extracting unit 364, a feature value setting unit 366, and a line defect determining unit 368. Yes. The line defect detection unit 360 extracts line defect candidates from the image to be inspected. Thereafter, it is determined whether or not the line defect candidate extracted by the line defect detection unit 360 is a line defect. The line defect detection unit 360 will be described later together with the flow of the line defect detection process.

  The output unit 400 outputs a determination result. Examples of the output unit 400 include a display unit such as a monitor and a printer.

  Next, the overall flow of the line defect detection process will be described with reference to FIG. FIG. 2 is a flowchart showing the flow of the line defect detection process of the present embodiment.

  First, the imaging means 200 images the screen 110 (S101). The inspection target region extraction unit 320 extracts the inspection target region from the obtained image information (S102), and then the background image difference processing unit 330 performs shading correction on the image information of the inspection target region (S103). Next, after the smoothing processing unit 340 removes noise components using a spatial filter (S104), the image duplicating unit 350 detects the type of line defect to be detected (black line defect extending in the X-axis direction). The image information is duplicated by the number corresponding to the white line defect extending in the X axis direction, the black line defect extending in the Y axis direction, and the white line defect extending in the Y axis direction (S105). Based on the image information obtained in this way, the line defect detection unit 360 creates a feature value map by assigning feature values to all the target pixels in the inspection target region in a predetermined procedure (S106). Based on this feature value map, line defect discrimination processing is performed (S107).

The detailed procedure of the processing shown in S106 and S107 will be described below.
The line defect detection process will be described by taking a detection process of a white line defect (hereinafter also referred to as a Y white line defect) extending in the Y-axis direction as an example.

  First, S106 will be described with reference to FIG. FIG. 3 is a flowchart showing a detailed procedure of S106 in FIG.

  The sampling area defining unit 362 first determines the size (size) of the sampling area (S111). The size of the sampling area is appropriately set according to the detection accuracy, the length of the line defect to be detected, the width of the line defect to be detected, the detection time, and the like. The shape of the sampling area may be a square (n × n) or a rectangle (n × m). In addition, n and m are, for example, 3 <n <50 and 3 <m <50, and it is desirable that n = 7 or 9 is a square. The reason why the odd number is used is that the position of the sampling region is determined so that the pixel of interest becomes the center of the sampling region in a later step. Therefore, if the pixel of interest is not located at the center of the sampling region, it may not be an odd number. In the case of a rectangle, it is preferable that the orientation of the sampling region is determined so that m is the length direction of the line defect when n <m. If n and m are too large, the processing time tends to increase. In this example, a 9 × 9 square sampling area is used.

  Next, as shown in FIG. 4, the sampling area defining unit 362 selects an arbitrary target pixel 116 from the inspection target area 112 (S112), and the sampling area 114 is centered on the target pixel 116. A position is defined (S113).

  The feature pixel extraction unit 364 sequentially scans the columns along the X axis in the sampling region 114, compares the luminance value of the pixel of the X axis coordinate value for each pixel column, and determines the pixel having the maximum luminance. The coordinate values (Max0 to Max8 in FIG. 5) are extracted (S114). Since it is assumed that there is only one Y white line defect in the minute sampling area 114 as in the present embodiment, by extracting the pixel having the maximum luminance for each column in this way, the Y white line defect is obtained. It is possible to predict where there is. This process is repeated until the investigation is completed for all the pixels in the sampling area 114 (S115).

  When the survey for all the pixels in the sampling area 114 is completed, the virtual line creating means (not shown) in the feature value setting unit 366 creates a virtual line from the coordinate value of the maximum luminance (S116).

  Here, a specific method of creating a virtual line will be described with reference to FIG. FIGS. 6A to 6C are diagrams for explaining a method of creating a virtual line. Here, the virtual line is obtained by calculating the first moment. First, assuming that there is a line defect in any one of the columns parallel to the Y axis in the sampling region 114, the sum of the distance moments (M0 to M8) between the straight line and each pixel having the maximum luminance ( Msi) is determined. Specifically, the virtual lines are sequentially moved in the X-axis direction, and the sum (Msi) of the distance moments with the pixels having the maximum luminance is obtained for all the columns (see FIGS. 6A and 6B). Here, the distance moment is the square of the distance between the X coordinate value of each pixel having the maximum luminance and the X coordinate value of the virtual line candidates (Fc0 to Fc8). Table 1 shows the result of calculating the sum of the distance moments between all the virtual line candidates in the sampling area and the pixel having the maximum luminance.

  In this table, an imaginary line candidate (here, Ms6) having the smallest sum (Ms0 to Ms8) of the distance moments (M0 to M8) is a phantom line. If there is a Y white line defect in this sampling area, it is considered that the Y white line defect exists near the coordinate value having the maximum luminance. Therefore, by performing such processing, a position with a high probability of the line defect is predicted. It becomes possible.

  Next, a distance calculation unit (not shown) in the feature value setting unit 366 calculates a distance (L) between the target pixel 116 and the virtual line Fc6 (S117, see FIG. 6C).

  The distance (L) between the virtual line Fc6 and the pixel of interest 116 is compared with an arbitrary distance threshold (Ls) set in advance (S118). If the distance is equal to or smaller than the threshold (Ls), a score is added to the pixel of interest 116. (Set characteristic value) (S119). If the distance exceeds the threshold (Ls), no score is added to the pixel of interest 116.

  In addition to the distance (L) calculated as described above, the score given to the target pixel 116 is a score based on the sum of distance moments (Msi, which is Ms6 in this example), the defect luminance difference (D), and the like. Is added. The defect luminance difference may be, for example, a difference between an average value of each maximum luminance and an average value of pixel luminances other than each maximum luminance value.

  As an example of the method of adding the score P1 (x, y), the following equation is given.

  The above process is repeated for all the pixels in the entire region to be inspected (the entire region), and a score is assigned to all the pixels (S120).

  The map creation means (not shown) in the feature value setting unit 366 creates a feature value map in which the points obtained for all the pixels as described above correspond to the coordinate values of the pixels (S121). FIG. 9 is a diagram illustrating an example of a feature value map. Note that a line defect detection image for determining the presence or absence of a line defect may be reconstructed based on the feature value map.

Based on the feature value map as described above, the line defect determination unit 368 determines a line defect.
S107 will be described with reference to FIG. FIG. 7 is a flowchart showing a detailed procedure of S107 in FIG.

First, an image (screen) size is acquired from the feature value map (S211).
Next, the feature value map is divided in the Y-axis direction.

  Specifically, first, the number of divisions in the Y-axis direction of the feature value map is determined, and the division size of the feature value map is calculated based on the acquired image size (S212). For example, when dividing into four in the Y-axis direction, when the size of the feature value map (entire image) is (x, y) = (1280, 960), the Y-axis divided size is (x, y) = (1280, 240). The division size is not particularly limited, but is appropriately set according to, for example, the length of the line defect to be detected.

  Next, the feature value map is divided in the Y-axis direction with a predetermined division size determined as described above (S213). At this time, a total of 2 × −1 divided areas divided by a predetermined division size from the edge of the feature value map and areas divided by a half pitch and divided to the predetermined division size (x: division when dividing from the image edge) Number) of divided areas.

  A method of dividing the feature value map will be described with reference to FIG. FIG. 8 is a diagram for explaining a feature value map dividing method. In FIG. 8, images corresponding to the feature value map are schematically shown for easy understanding. As shown in FIG. 8, when the line defect detection image is divided into four, four divided regions (FIG. 8 (b)) divided into four from the pre-division image (feature value map) (see FIG. 8 (a)). )) And three divided areas shifted by half pitch (see FIG. 8C) are obtained, so that a total of seven divided areas are obtained.

These divided regions are integrated in the dividing direction (Y-axis direction), respectively (S214). The maximum integrated value is calculated (S215). By repeating this process, the maximum value of the integrated values is calculated for all the divided areas (S216). Next, the largest value among the maximum values acquired for each divided region is extracted (S217). The acquired maximum value is compared with an arbitrary determination threshold value NG _th (threshold value serving as a determination criterion) (S218). If the value is equal to or greater than the determination threshold value, it is determined that there is a line defect (defective product) (S219). If it is less than the determination threshold, it is determined that there is no line defect (non-defective product). The presence or absence of a line defect is determined by such processing.

  Hereinafter, a specific example will be described. For example, when the maximum integrated value of each divided region obtained from the divided image of the line defect detection image shown in FIG. 10A is as shown in Table 2 below, the divided region having the largest integrated value is the largest. No. 5

Here, no. The divided image of 5 is judged to have a line defect (defective product) because NG _th <2485 when the determination threshold value NG _th = 1000.

  When the maximum value obtained from the divided image of the line defect detection image shown in FIG. 10B is as shown in Table 3 below, the divided image having the largest integrated value is No. 3

Here, no. In the divided image 3, when the determination threshold value NG _th = 1000, NG _th > 512, it is determined that there is no line defect (non-defective product).

  The inspection results based on the line defect detection method of the present embodiment are shown in FIGS. FIG. 11A is a diagram for referring to the position of a line defect, and FIG. 11B is a diagram illustrating a detection result when a line defect that is easy to detect is detected by a conventional method. (C) is a figure which shows the detection result at the time of detecting by the method of this embodiment. Here, the conventional method is data obtained by integrating the screen data in the Y-axis direction without performing the processing as in the present embodiment. As shown in FIG. 11 (c), in the conventional method, the peak corresponding to the line defect is unclear, whereas in the present invention, the peak of the line defect stands out. 12A is a diagram for referring to the position of the line defect, and FIG. 12B is a diagram illustrating a detection result when a line defect at a visual limit is detected by the conventional method. (C) is a figure which shows the detection result at the time of detecting by the method of this embodiment. As can be seen from FIGS. 12B and 12C, according to the method of the present embodiment, it is possible to easily detect a line defect at a visual limit level that is difficult to detect by the conventional method.

  According to the present embodiment, feature pixels that are candidates for line defects (also referred to as linear defects) with the maximum or minimum luminance are extracted for each pixel column, and features based on distance information with respect to the feature pixels for all pixels. By assigning a value, the effect of point defects other than line defects (also called point defects) and noise can be reduced, so that line defects can be detected with high sensitivity, and the detection results obtained High reliability. In addition, by creating a virtual line from the feature pixel by calculating the first moment, it is possible to create a virtual line at a position where there is a high probability of a line defect. Can be set). This further improves the reliability of line defect detection.

  In the above example, an example of creating a feature value map as an intermediate process has been described. However, when a CPU having a sufficiently high processing speed is used, direct integration processing is performed without creating a feature value map. Also good.

  In the above example, in the line defect determination process, the presence / absence of a line defect is determined based on the maximum integrated value. However, the present invention is not limited to this. For example, the difference between the maximum integrated value and the average integrated value is determined. May be determined by comparing with a predetermined threshold.

  In the above example, the target pixel is set to be approximately in the center of the sampling area. However, the present invention is not limited to this. For example, the target pixel may be set to be arranged at an end in the sampling area.

  In the above example, the method for detecting a white line defect has been described. However, a black line defect can be detected by detecting a pixel having the minimum luminance instead of detecting a pixel having the maximum luminance. In the above example, the method for detecting a line defect extending on the Y-axis has been described. However, instead of this, the pixel detection direction, the division direction of the line defect detection image, and the like can be changed as appropriate. It is also possible to detect a line defect extending on the X axis by the above process. Therefore, by detecting the number of images corresponding to the type of line defect in the image duplicating unit 350, it is possible to simultaneously detect such other line defects.

(Modification)
Next, a modified example of the sampling area defining method will be described.
FIG. 13 is a diagram for explaining a modification of the sampling area defining method.

  As shown in FIG. 13A, when the sampling area 114 is defined so that the pixel of interest is arranged at the center at the edge of the image, the sampling area 114 protrudes outside the image area. In such a case, in this example, as shown in FIG. 13B, the sampling area 114 is moved vertically and horizontally so as to be within the image area. Specifically, the positions of the end of the sampling area 114 and the end of the image are matched. Thereafter, the same processing as usual can be performed.

  In this way, by aligning the edge of the sampling area with the edge of the image area, the sampling area can be accommodated in the image area, so there is no provision of a virtual area or setting a virtual value. Therefore, it becomes possible to give the pixel of interest distance information with more accurate line defects. Therefore, a more reliable detection result can be obtained.

(Second Embodiment)
In the first embodiment, a virtual line is extracted in the sampling area, and a score (feature value) is given to the target pixel based on the distance between the virtual line and the target pixel. However, the method for giving the score is limited to this. Is not to be done.

In this embodiment, another aspect of a method for assigning a score to a target pixel will be described.
FIG. 14 is a flowchart showing the flow of the line defect detection process of the present embodiment. The steps of S311 to S315 are the same as S111 to S115 in FIG.

  After extracting the feature pixels by the feature pixel extraction unit 364, in this embodiment, the distance calculation unit (not shown) included in the feature value setting unit 366 calculates the distance between each feature pixel and the pixel of interest in the x-axis direction, The sum is obtained (S316). This will be specifically described below.

First, the square (distance moment) of the difference between the x coordinate value of each feature pixel and the x coordinate value of the target pixel is calculated. FIG. 15 is a diagram for explaining a method of calculating the distance between each feature pixel and the target pixel in an arbitrary sampling region. Here, as shown in FIG. 15, from the uppermost row parallel to the x-axis direction, m ₀ = 2 ² = 4, m ₁ = 4 ² = 16, m ₂ = 1 ² = 1, m ₃ = 2 ² = 4, m ₄ = 3 ² = 9, m ₅ = 1 ² = 1, m ₆ = 0 ² = 0, m ₇ = 3 ² = 9, m ₈ = 2 ² = 4. Next, a sum m _si of m _{0 to} m ₈ is calculated. Taking FIG. 15 as an example, the sum m _si is as follows.

The feature value setting unit 366 compares the sum m _si calculated in each sampling region with the distance threshold m _th in this way to determine the presence or absence of a line defect candidate (S317). For example, when m _th = 50, in FIG. 16A, m _s1 = 9 <m _th and it is determined that there is a line defect candidate, and in FIG. 16B, m _s2 = 81> M _{th and the} line defect. It is determined that there are no candidates. In FIG. 16C, m _S3 = 16 <m _th and it is determined that there is a line defect candidate.

  When it is determined that there is a line defect candidate, the feature value setting unit 366 assigns a point (feature value) corresponding to the distance information to the pixel of interest in the sampling region determined to have the line defect candidate (S318). ). An example of the score giving formula is as follows.

  In the above formula, a and b are arbitrary variable coefficients (floating point type). The reason why the distance coefficient is +1 is to prevent the denominator from becoming 0 because the minimum value of the distance coefficient is 0. The reason why the luminance difference is an absolute value is to prevent the luminance difference from becoming a negative value because the luminance difference value is positive or negative depending on whether it is bright or dark with respect to the background.

  Thereafter, the above process is repeated for all pixels to form a feature value map (S319, S320). Next, the presence / absence of a line defect can be determined by performing a determination process similar to that of the first embodiment.

  In the present embodiment, since the feature value is set for the pixel of interest based on the sum of the squares of the distances, it is possible to set a feature value that is relatively simple and reliable. Therefore, it is possible to obtain a detection result that is fast and reliable.

(Third embodiment)
When the line width of the line defect of the captured image is large and the line width becomes larger than the width of one pixel, the detection may be impossible. Even in such a case, in the present embodiment, a process of reducing an image to be inspected is performed before the line defect detection process so that the line defect can be detected.

Hereinafter, the present embodiment will be described with reference to FIGS.
FIG. 17 is a flowchart showing a flow of line defect detection processing in the present embodiment, and FIG. 18 is a flowchart showing a detailed procedure of S405 in FIG.

  This embodiment is the same as the first embodiment except that the image reduction process (S405) is performed before the image duplication process (S406). In S405, a process of reducing only the direction component perpendicular to the extending direction of the line defect to be detected in the image is performed.

  The specific flow of the image reduction process will be described below with reference to FIG. In the present embodiment, a case where a Y-ray defect is detected will be described as an example, but the present invention is not limited to this.

  First, when the image reduction process is started, the image reduction ratio in the X-axis direction is set (S411). Here, an appropriate value may be determined in advance as the reduction ratio. In addition, a reduction ratio corresponding to the thickness of the line defect assumed to be detected this time may be appropriately determined depending on the case. Further, it is preferable to determine at least two reduction ratios (eg, reduction ratios 1/2 and 1/4, etc.) so as to deal with line defects having various line widths. Next, the number of images corresponding to the type of reduction ratio is duplicated (S412). Next, each of the obtained duplicate images is reduced to a preset reduction rate (S413).

  FIG. 19A is a diagram showing the original image size, FIG. 19B is an image when the reduction ratio is 1/2, and FIG. 4 is an image when 4. As shown in FIGS. 19A to 19C, by reducing the image in the direction in which the line width of the line defect is reduced, for example, the line defect that is too thick and cannot be detected with the original image as it is When P2 and P3 are reduced by 1/2, P2 can be detected first, and further, when P3 and P3 are reduced by 1/4, P3 can be detected. That is, by appropriately changing the reduction ratio, it is possible to detect a line defect that could not be detected with the original image size according to the present embodiment. Also, by obtaining images with a plurality of reduction ratios at the same time, it is possible to simultaneously perform line defect detection processing in parallel, so that it is possible to obtain a detection result with higher reliability at a higher speed.

FIG. 1A is a configuration diagram showing an example of a screen line defect detection apparatus of the present invention, and FIG. 1B is a diagram for explaining a detailed configuration of a line defect detection unit. FIG. 2 is a flowchart showing the flow of the line defect detection process of the present embodiment. FIG. 3 is a flowchart showing a detailed procedure of S106 in FIG. FIG. 4 is a diagram for explaining the relationship between the pixel of interest and the sampling region. FIG. 5 is a diagram for explaining the feature value extraction processing. FIG. 6 is a diagram for explaining a method of creating a virtual line. FIG. 7 is a flowchart showing a detailed procedure of S107 in FIG. FIG. 8 is a diagram for explaining a feature value map dividing method. FIG. 9 is a diagram illustrating an example of a feature value map. FIG. 10 is a diagram illustrating a specific example of a line defect. FIG. 11 is a diagram illustrating a difference in detection sensitivity of the line defect between the first embodiment and the conventional method. FIG. 12 is a diagram showing a difference in detection sensitivity of line defects between the first embodiment and the conventional method. FIG. 13 is a diagram for explaining a modification of the sampling area defining method. FIG. 14 is a flowchart illustrating a flow of line defect detection processing according to the second embodiment. FIG. 15 is a diagram for explaining a method of calculating the distance between each feature pixel and the target pixel in an arbitrary sampling region. FIG. 16 is a diagram for explaining a feature value setting method according to the second embodiment. FIG. 17 is a flowchart showing the flow of line defect detection processing in the third embodiment. FIG. 18 is a flowchart showing a detailed procedure of S405 in FIG. FIG. 19A is a diagram showing the original image size, FIG. 19B is an image when the reduction ratio is 1/2, and FIG. 4 is an image when 4.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 ... Projector, 110 ... Screen, 112 ... Inspection object area, 114 ... Sampling area, 116 ... Pixel of interest, 120 ... Screen, 200 ... Imaging means, 300 ... Image information processing unit, 310 ... input unit, 320 ... inspection target region extraction unit, 330 ... background image difference processing unit, 340 ... smoothing processing unit, 350 ... image replication unit, 360... Line defect detector, 362... Sampling area demarcater, 364... Feature pixel extractor, 366... Feature value setting unit, 368. Part

Claims (15)

A sampling region defining step of selecting a target pixel from the image information of the input screen and demarcating the sampling region so as to include the target pixel;
A feature pixel extraction process for extracting a feature pixel having a maximum or minimum luminance for each pixel column from the pixel data of the sampling region;
A feature value setting process for setting a value obtained by weighting the target pixel according to the distance between the predicted position of the line defect obtained from the position information of the feature pixel and the target pixel;
By repeating the process from the sampling area defining process to the feature value setting process for each pixel of interest, a feature value is set for all the pixels in the entire area of the screen, and the line defect of the screen is based on the feature value. A discriminating process for discriminating
A method for detecting a line defect on a screen. The feature value setting process includes:
Creating a virtual line from the feature pixels;
Calculating a distance between the virtual line and the pixel of interest;
A process of setting a weighted value according to the distance to the pixel of interest;
The method for detecting a line defect according to claim 1, comprising: The feature value setting process includes:
A step of calculating a distance between a reference line that includes the target pixel and is perpendicular to the pixel column, and the feature pixels;
A process of setting a weighted value according to the distance to the pixel of interest;
The method for detecting a line defect according to claim 1, comprising: The discrimination process is
A process of creating a feature value map in which feature values are set for all pixels in the entire area by repeatedly performing the process from the sampling area defining process to the feature value setting process for each pixel of interest;
Determining a line defect on the screen based on the feature value map;
The method for detecting a line defect according to claim 1, comprising:   The process of determining line defects on the screen is a process of detecting line defects by integrating the feature values based on the feature value map and comparing the maximum value of the integrated values with a predetermined threshold value. The method for detecting a line defect according to claim 4. The process of determining line defects on the screen
A division process of dividing the feature value map in a uniaxial direction;
For each of the divided areas, a maximum value calculation process for calculating the maximum value of the integrated values by integrating the feature values;
A line defect detection process for detecting a line defect by comparing the maximum value of the integrated values of each of the divided regions and comparing the maximum value among the maximum values with a predetermined threshold value;
The method for detecting a line defect according to claim 4, comprising: The process of determining line defects on the screen
A first maximum value calculating step of determining a predetermined integration range in the feature value map, integrating the feature values within the integration range and obtaining a maximum value of the integration values;
A second maximum value calculating process for setting the next integrated range so as to partially overlap the previous integrated range, and obtaining the maximum value of the integrated value of the feature value for the next integrated range;
By repeating the second maximum value calculation process, integration is performed over the entire range in the feature value map, the maximum value of the integrated values of each integrated range is compared, and the largest value among the maximum values is determined. A line defect detection process for detecting line defects by comparing with a predetermined threshold;
The method for detecting a line defect according to claim 4, comprising: Before the sampling area defining step, the image information is compressed in a uniaxial direction to obtain a compressed image,
The method for detecting a line defect according to claim 1.   9. The line defect detection method according to claim 8, wherein the image compression process is a process of compressing the image information at at least two compression ratios to obtain at least two compressed images. The sampling area defining process includes:
When the sampling area is determined so that the pixel of interest is normally located in the center and the pixel of interest is determined to be located in the approximate center of the sampling area, the sampling area protrudes from the entire area to be inspected. The position of the sampling region is determined so that the end of the sampling region coincides with the end of the entire region so that the sampling region fits in the entire region only. Line defect detection method.   The line defect detection method according to claim 1, further comprising an extraction step of extracting a region to be inspected from input image information on the screen before the sampling region defining step.   Sampling area demarcating means for selecting a target pixel from the image information of the input screen and demarcating the sampling area so as to include the target pixel, and the luminance is maximum or minimum for each pixel column from the pixel data of the sampling area A feature pixel extracting unit that extracts a feature pixel, and a feature that sets a weighted value according to a distance between an expected position of a line defect obtained from position information of the feature pixel and the target pixel to the target pixel A screen including a line defect detection unit configured to set a feature value for each pixel of interest in the entire area of the screen and determine a line defect on the screen based on the feature value. Line defect detection device. The feature value setting means is
A virtual line creating unit that creates a virtual line from the feature pixel, a distance calculating unit that calculates a distance between the virtual line and the pixel of interest, and a value obtained by weighting the pixel of interest according to the distance are set. Setting means;
The line defect detection device according to claim 12, comprising: The feature value setting means is
A reference line that includes the target pixel and is perpendicular to the pixel column, a distance calculation unit that calculates the distance between each feature pixel, and a setting that sets a value that is weighted according to the distance to the target pixel. Means,
The line defect detection device according to claim 12, comprising: The line defect detection device according to claim 12, wherein the determination unit creates a feature value map based on the feature value after setting a feature value for each pixel of interest in the entire region.

Images