A display panel ripple defect detection method and device

Abstract

The application discloses a display panel wave defect detection method and device, and relates to the technical field of display panel comprehensive testing, and the method comprises the following steps: performing tilt correction and distortion correction on a region of interest (ROI) of a display panel wave image; performing enhancement processing on a corrected image after the tilt correction and the distortion correction; calculating position information of all defects in the ROI after the enhancement processing; and setting a periodically changing minimum pixel distance to measure adjacent defects, so as to periodically determine the defects. The application can reduce detection cost and improve detection efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of display panel automatic optical detection, in particular to a display panel ripple defect detection method and device. BACKGROUND

[0002] The surface of a watch display panel generates a non-uniform image under an SGT (Shadowgraph tester) device, and the flatness (or ripple) of the target image surface is inconsistent, which may cause multiple specific types of ripple defects. The WSGT (Watch Shadowgraph tester) project is to identify and classify the ripple defects and output the intensity scores of various defects.

[0003] The existing commonly used ripple detection is mostly in the form of deep learning target detection, which requires a large number of data sets, has low segmentation accuracy, and has high maintenance and debugging costs. There are many types of defects, and sample labeling requires a lot of time, and the overall development cycle is long. SUMMARY

[0004] In view of the defects in the prior art, the first aspect of the present application provides a display panel ripple defect detection method, which can reduce the detection cost and improve the detection efficiency.

[0005] To achieve the above purpose, the technical solution adopted by the present application is:

[0006] A display panel ripple defect detection method, the method comprising the following steps:

[0007] The region of interest (ROI) of the display panel ripple image is subjected to tilt correction and distortion correction;

[0008] The corrected image after tilt correction and distortion correction is subjected to enhancement processing;

[0009] The position information of all defects in the ROI after enhancement processing is calculated, and the minimum pixel distance between adjacent defects is set to periodically change, so as to periodically determine the defects.

[0010] In some embodiments, the calculation of the position information of all defects in the ROI after enhancement processing, and the setting of the minimum pixel distance between adjacent defects to periodically change, so as to periodically determine the defects, comprises:

[0011] Determine the horizontal and vertical coordinates and corresponding serial numbers of all defects in the ROI after enhancement processing;

[0012] Calculate the difference value of the horizontal and vertical coordinates of adjacent defects, and determine a serial number set A that satisfies the horizontal coordinate difference value being greater than the average width of the defects and the vertical coordinate difference value being greater than the average height of the defects;

[0013] a minimum pixel distance D is set, and a local peak value set B is determined, in which the difference value of the horizontal and vertical coordinates is greater than D;

[0014] Discrete defects belonging to the set B are deleted from the set A.

[0015] In some embodiments, the method further comprises a step of calculating the inner and outer contrast of the defects, comprising:

[0016] The ROI after the enhancement processing is converted into a first defect binary image BW1;

[0017] The center coordinates and inner ring contrast of each defect in BW1 are calculated.

[0018] A disc-shaped structure element is created to perform an inflation operation on BW1 to obtain a second defect binary image BW2;

[0019] The center coordinates and reference contrast of each defect in BW2 are calculated.

[0020] The distance between each defect in BW2 and the center coordinates of all defects in BW1 is calculated in sequence.

[0021] The reference contrast of the defect with the smallest distance in BW2 is selected as the outer ring contrast of a defect in BW1.

[0022] In some embodiments, the enhancement processing on the corrected image after the tilt correction and distortion correction comprises:

[0023] The corrected image is subjected to Gaussian filtering, and a difference image of Gaussian filtering under different variances is calculated, and the negative part of the difference image is taken as a first background mask and the positive part is taken as a foreground;

[0024] The corrected image is subjected to foreground and background segmentation according to the maximum inter-class variance method to obtain a second background mask;

[0025] The corrected image is subjected to fixed threshold segmentation according to the binary method to obtain a third background mask;

[0026] The product of the difference between the first and second background masks and the third background mask is calculated to obtain an actual background mask of the corrected image, and the brightness mean value of the actual background intensity region of the corrected image is determined.

[0027] A ratio coefficient of the set fixed background intensity and the mean value of the actual background intensity of the corrected image is calculated, and the foreground is stretched based on the ratio coefficient to obtain an enhanced image.

[0028] In some embodiments, the tilt correction on the region of interest (ROI) of the display panel ripple image comprises:

[0029] After performing binary segmentation of the ROI, the Canny operator is used to calculate the ROI boundary.

[0030] Set a tilted mask with an angle equal to the diagonal of the ROI;

[0031] Calculate the intersection of the ROI boundary and the tilted mask to segment out the diagonal boundary;

[0032] The angles of each side are calculated using the least squares method, and the mean angle is used for tilt correction.

[0033] In some embodiments, the distortion correction of the region of interest (ROI) of the display panel ripple image includes:

[0034] After performing image binary segmentation on the tilt-corrected ROI, the ROI boundary is recalculated using the Canny operator.

[0035] Set up a horizontal mask and an angled rectangular mask;

[0036] Calculate the intersection of the ROI boundary with the horizontal mask and the rectangular mask to segment the horizontal and vertical boundaries;

[0037] Calculate the first-order linear functions of each horizontal and vertical boundary using the least squares method, and calculate the first intersection point between each pair of boundaries;

[0038] The average of the x-coordinates of any first intersection point and another first intersection point on the horizontal boundary where the first intersection point is located, and the average of the y-coordinates of another first intersection point on the vertical boundary where the first intersection point is located, are taken as the second intersection point of the ROI boundary.

[0039] Calculate the perspective transformation matrix of the first and second intersection points to perform distortion correction.

[0040] In some embodiments, the rectangular mask is tilted at an angle of 75°.

[0041] A second aspect of the present invention provides a display panel ripple defect detection device, which can reduce detection costs and improve detection efficiency.

[0042] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0043] A display panel ripple defect detection device, comprising:

[0044] The correction module is used to perform tilt and distortion correction on the region of interest (ROI) of the display panel ripple image.

[0045] The image enhancement module is used to enhance the corrected image after tilt correction and distortion correction.

[0046] A determination module is configured to calculate the position information of all defects in the ROI after the enhancement processing, and set a periodically changed minimum pixel distance D to measure adjacent defects to periodically determine the defects.

[0047] In some embodiments, the determination module is configured to calculate the position information of all defects in the ROI after the enhancement processing, and set a periodically changed minimum pixel distance D to measure adjacent defects to periodically determine the defects, including:

[0048] determining the horizontal and vertical coordinates and corresponding serial numbers of all defects in the ROI after the enhancement processing;

[0049] calculating the difference values of the horizontal and vertical coordinates of adjacent defects, and determining a serial number set A satisfying the condition that the horizontal coordinate difference value is greater than the average width of the defects, and the vertical coordinate difference value is greater than the average height of the defects;

[0050] setting a periodically changed minimum pixel distance D, and determining a local peak serial number set B with a minimum height difference greater than D;

[0051] removing discrete defects belonging to the serial number set B from the serial number set A.

[0052] In some embodiments, the method further comprises a calculation module configured to calculate the inner and outer contrasts of the defects, and the calculation module is configured to:

[0053] convert the ROI after the enhancement processing into a first defect binary image BW1;

[0054] calculate the center coordinates and inner ring contrasts of each defect in BW1;

[0055] create a disc-shaped structure element to perform an inflation operation on BW1 to obtain a second defect binary image BW2;

[0056] calculate the center coordinates and reference contrasts of each defect in BW2;

[0057] calculate the distances between each defect in BW2 and the center coordinates of all defects in BW1 in sequence;

[0058] select the reference contrast of the defect with the minimum distance in BW2 as the outer ring contrast of a defect in BW1.

[0059] Compared with the prior art, the present application has the following advantages:

[0060] The display panel wave defect detection method in the application improves the tilt correction, distortion correction, image enhancement, defect determination and defect inside and outside contrast in the defect detection process. Especially the periodic defect determination, which can screen and classify different defects by setting the periodically changing minimum pixel distance. Compared with the deep learning target detection method in the prior art, which has low segmentation accuracy, high maintenance and debugging cost, and requires a lot of time to mark samples due to the large number of defect types, the application can classify defects according to different shapes, sizes, positions, contrasts and other characteristics, and finally score the intensity of each type of defect to measure the quality level of the display panel, thereby reducing the detection cost and improving the detection efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0061] Figure 1 is a flowchart of the display panel wave defect detection method in the embodiment of the application;

[0062] Figure 2 is a schematic diagram of the correction process in the embodiment of the application;

[0063] Figure 3 is a schematic diagram of image enhancement and segmentation screening in the embodiment of the application;

[0064] Figure 4 is a schematic diagram of the tilt correction in the embodiment of the application;

[0065] Figure 5 is a schematic diagram of the distortion correction in the embodiment of the application. DETAILED DESCRIPTION

[0066] To make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0067] Referring to Figures 1 to 3 The embodiment of the application provides a display panel wave defect detection method, which comprises the following steps:

[0068] S1. Perform tilt correction and distortion correction on the region of interest ROI (Region of Interest) of the display panel wave image.

[0069] In the embodiment of the present application, the watch display panel wave image is taken as an example for illustration. Specifically, since the ROI of the watch display panel is in a tilted state, the tilt and distortion of the ROI are respectively corrected by calculating the horizontal angle and the perspective transformation matrix. In this embodiment, the tilt correction is performed first, and then the distortion correction is performed.

[0070] Referring to Figure 4 shown, for tilt correction:

[0071] First, the ROI is subjected to image binary segmentation, and then the Canny operator is used to calculate the ROI boundary; then a tilt mask equal to the diagonal line angle of the ROI is set; then the intersection of the ROI boundary and the tilt mask is calculated to segment the diagonal boundary; finally, the least square method is used to calculate the angle of each side, and the tilt correction is performed using the average angle.

[0072] Referring to Figure 5 shown, for distortion correction:

[0073] First, the ROI after tilt correction is subjected to image binary segmentation, and then the Canny operator is used to calculate the ROI boundary again; then a horizontal direction mask and an angle-based rectangular mask are set; then the intersection of the ROI boundary and the horizontal direction mask and the rectangular mask is calculated to segment the horizontal direction boundary and the vertical direction boundary; then the least square method is used to calculate the first linear function of each horizontal direction boundary and vertical direction boundary, and the first intersection point between each other is calculated; then, the second intersection point of the ROI boundary is taken as the average value of the horizontal direction boundary of the first intersection point and the other first intersection point, and the average value of the vertical direction boundary of the first intersection point and the other first intersection point; finally, the perspective transformation matrix of the first intersection point and the second intersection point is calculated to perform distortion correction.

[0074] For the convenience of understanding the calculation method of the second intersection point, referring to the parallelogram in Figure 5 , the four vertices of which are A(x1, y1), B(x2, y2), C(x3, y3), and D(x4, y4), wherein A is the upper left corner, and ABCD is in the counterclockwise direction.

[0075] Then, the four vertices (second intersection points) corresponding to the rectangle are:

[0076] A'((x1+x2) / 2, (y1+y4) / 2); B'((x1+x2) / 2, (y2+y3) / 2);

[0077] C'((x3+x4) / 2, (y2+y3) / 2); D'((x3+x4) / 2, (y1+y4) / 2).

[0078] S2. Perform enhancement processing on the corrected image after tilt correction and distortion correction.

[0079] Specifically, step S2 includes:

[0080] S21. Perform Gaussian filtering on the corrected image according to the formula:

[0081]

[0082]

[0083] Calculate the difference image DOG of Gaussian filtering under different variances δ, and take the negative part of the difference image DOG as the first background mask MaskDOG and the positive part as the foreground Δ;

[0084] S22. Perform foreground and background segmentation on the corrected image according to the maximum inter-class variance method to obtain a second background mask Mask OTSU ;

[0085] S23. Perform fixed threshold segmentation on the corrected image according to the binary method to obtain a third background mask Mask TM ;

[0086] According to the formula:

[0087] Mask BG = (Mask DOG - Mask OTSU ) * Mask TM

[0088]

[0089] Calculate the ratio coefficient of the set fixed background intensity Intensity and the average of the actual background intensity of the corrected image, where Mask BG is the actual background mask of the corrected image, and GMean(Mask BG ) is the average brightness of the actual background MaskBG region of the corrected image;

[0090] S24. Stretch the foreground Δ based on Mult to obtain an enhanced image;

[0091] In step S24, the foreground Δ is stretched to obtain the final enhanced image Δ', which can reduce the segmentation error caused by the difference in image brightness. That is, Δ' = Mult × Δ.

[0092] S3. Calculate the position information of all defects in the ROI after enhancement processing, and set a periodically changing minimum pixel distance to measure adjacent defects to periodically determine the defects.

[0093] Specifically, step S3 includes:

[0094] S31. Calculate the position information of all defects in the ROI after enhancement processing;

[0095] S32. Sort the horizontal and vertical coordinates of the defects to obtain the position and sequence number: [x, Indx] and [y, Indy];

[0096] S33. Calculate the difference dx and dy of adjacent defects;

[0097] S34. Calculate the average width xmean and average height ymean of the defects, and calculate the sequence number set A of dx>xmean and dy>ymean;

[0098] S35. Set the periodically changing minimum pixel distance D, and determine the local peak sequence number set B of the minimum height difference of dx and dy greater than D;

[0099] S36. Delete the defects corresponding to Indx∈A-B and Indy∈A-B.

[0100] It can be understood that after setting the periodically changing minimum pixel distance D, since the minimum pixel distance of each period is different, the elements in set B will also be different, which is equivalent to classifying the size and position of the defects in each period, and screening the discrete defects through step S36.

[0101] In some preferred embodiments, the step of calculating the inner and outer contrasts of the defects is also included, which comprises:

[0102] Convert the ROI after enhancement processing into a first defect binary image BW1;

[0103] Calculate the center coordinates and inner ring contrast of each defect in BW1;

[0104] Create a disc-shaped structure element with a radius of r to perform dilation operation on BW1 to obtain a second defect binary image BW2;

[0105] Calculate the center coordinates and reference contrast of each defect in BW2;

[0106] Calculate the distance between each defect in BW2 and the center coordinates of all defects in BW1 in turn;

[0107] Select the reference contrast of the defect with the smallest distance in BW2 as the outer ring contrast of a defect in BW1.

[0108] In summary, the display panel ripple defect detection method in the application improves the tilt correction, distortion correction, image enhancement, defect determination and defect inner-outer contrast in the defect detection process. Especially for the periodic defect determination, the different defects can be screened and classified by setting the periodically changing minimum pixel distance. Compared with the deep learning target detection method in the prior art, which has low segmentation accuracy, high maintenance and debugging cost, and requires a lot of time for sample marking due to the large number of defect types, the application can classify defects according to different shapes, sizes, positions, contrasts and other characteristics, and finally score the intensity of each type of defect to measure the quality grade of the display panel, thereby reducing the detection cost and improving the detection efficiency.

[0109] Meanwhile, the application also provides a display panel ripple defect detection device, which comprises a correction module, an image enhancement module and a determination module.

[0110] The correction module is used for tilt correction and distortion correction of a region of interest (ROI) of a display panel ripple image; the image enhancement module is used for enhancement processing of the corrected image after tilt correction and distortion correction; and the determination module is used for calculating position information of all defects in the ROI after enhancement processing, setting a periodically changing minimum pixel distance to measure adjacent defects, and performing periodic determination of the defects.

[0111] In some embodiments, the determination module calculates the position information of all defects in the ROI after enhancement processing, sets a periodically changing minimum pixel distance to measure adjacent defects, and performs periodic determination of the defects, including:

[0112] calculating the position information of all defects in the ROI after enhancement processing;

[0113] sorting the horizontal coordinates and vertical coordinates of the defects to obtain the position and serial number: [x, Indx] and [y, Indy];

[0114] calculating the difference dx and dy of adjacent defects;

[0115] calculating the average width xmean and average height ymean of the defects, and calculating the serial number set A of dx>xmean and dy>ymean;

[0116] setting a periodically changing minimum pixel distance D, determining the local peak serial number set B of the minimum height difference dx and dy greater than D;

[0117] deleting the defects corresponding to Indx∈A-B and Indy∈A-B.

[0118] In some embodiments, the application further comprises a calculation module for calculating the inner-outer contrast of the defects, and the calculation module is used for:

[0119] Convert the enhanced ROI to a first defect binary image BW1;

[0120] Calculate the center coordinates and inner ring contrast of each defect in BW1;

[0121] Create a disc-shaped structure element to perform dilation operation on BW1 to obtain a second defect binary image BW2;

[0122] Calculate the center coordinates and reference contrast of each defect in BW2;

[0123] Calculate the distance between each defect in BW2 and the center coordinates of all defects in BW1 in turn;

[0124] Select the reference contrast of the defect with the smallest distance in BW2 as the outer ring contrast of a defect in BW1.

[0125] In some embodiments, the image enhancement module performs enhancement processing on the corrected image after tilt correction and distortion correction, including:

[0126] Gaussian filtering is performed on the corrected image according to the formula:

[0127]

[0128]

[0129] Calculate the difference image DOG of Gaussian filtering under different variances δ, and take the negative part of the difference image DOG as the first background mask MaskDOG and the positive part as the foreground Δ;

[0130] According to the maximum inter-class variance method, the corrected image is segmented into foreground and background to obtain a second background mask Mask OTSU ;

[0131] According to the binary method, the corrected image is segmented by a fixed threshold to obtain a third background mask Mask TM ;

[0132] According to the formula:

[0133] Mask BG = (Mask DOG - Mask OTSU )*Mask TM

[0134]

[0135] Calculate the ratio coefficient of the set fixed background intensity Intensity and the average of the actual background intensity of the corrected image, where Mask BG is the actual background mask of the corrected image, and GMean(MaskBG ) to correct the brightness mean value of the actual background MaskBG region of the image;

[0136] Based on the Mult, the foreground Δ is stretched to obtain an enhanced image.

[0137] In some embodiments, the correction module performs tilt correction on a region of interest (ROI) of the display panel ripple image, including:

[0138] After image binary segmentation of the ROI, a Canny operator is used to calculate the ROI boundary;

[0139] A tilt mask with an angle equal to the diagonal line angle of the ROI is set;

[0140] The intersection of the ROI boundary and the tilt mask is calculated to segment the diagonal boundary;

[0141] According to the least squares method, the angle of each side is calculated, and the average angle is used for tilt correction.

[0142] In some embodiments, the correction module performs distortion correction on a region of interest (ROI) of the display panel ripple image, including:

[0143] After image binary segmentation of the tilt-corrected ROI, a Canny operator is used to calculate the ROI boundary again;

[0144] A horizontal mask and an angle-bearing rectangular mask are set;

[0145] The intersection of the ROI boundary and the horizontal mask and the rectangular mask is calculated to segment the horizontal boundary and the vertical boundary;

[0146] According to the least squares method, a first linear function of each horizontal boundary and vertical boundary is calculated, and the first intersection point between each two is calculated;

[0147] Taking any first intersection point, the average of the horizontal coordinate of another first intersection point of the horizontal boundary where the first intersection point is located, and the average of the vertical coordinate of another first intersection point of the vertical boundary where the first intersection point is located, as the second intersection point of the ROI boundary;

[0148] The perspective transformation matrix of the first intersection point and the second intersection point is calculated to perform distortion correction.

[0149] In some embodiments, the tilt angle of the rectangular mask is 75°.

[0150] The foregoing detailed description of the application has been presented for purposes of illustration and description. Various modifications and changes can be made to these embodiments without departing from the spirit and scope of the application. It is intended that the scope of the application should not be limited by the particular representative embodiments described above.

Claims

1. A method for detecting a moire defect of a display panel, characterized in that, The method includes the following steps: Tilting and distortion corrections are performed on the region of interest (ROI) of the display panel ripple image; Enhancement processing is applied to the corrected images after tilt and distortion correction; Calculate the location information of all defects within the enhanced ROI, and set a periodically changing minimum pixel distance metric for adjacent defects to periodically determine the defects. The calculation involves determining the location information of all defects within the enhanced ROI, setting a periodically varying minimum pixel distance metric for adjacent defects, and periodically determining the defects, including: Determine the x and y coordinates and corresponding numbers of all defects within the enhanced ROI; Calculate the difference between the horizontal and vertical coordinates of adjacent defects, and determine the set A of the indices that satisfy the condition that the difference between the horizontal coordinates is greater than the average width of the defect and the difference between the vertical coordinates is greater than the average height of the defect. Set a periodically changing minimum pixel distance D, and determine the set of local peak indices B where the minimum height difference between the horizontal and vertical coordinates is greater than D; Remove discrete defects belonging to set B from set A. 2.The method of claim 1, wherein, It also includes the step of calculating the internal and external contrast of the defect, including: The enhanced ROI is converted into a first defect binary map BW1; Calculate the center coordinates and inner circle contrast of each defect in BW1; Create a disk-shaped structural element and perform an expansion operation on BW1 to obtain the second defect binary image BW2; Calculate the center coordinates and reference contrast of each defect in BW2; Calculate the distance between the center coordinates of each defect in BW2 and the center coordinates of all defects in BW1 in turn; In BW2, the reference contrast of the defect with the smallest distance is selected as the outer ring contrast of a defect in BW1. 3.The method of claim 1, wherein, The enhancement processing of the corrected image after tilt correction and distortion correction includes: Gaussian filtering is applied to the corrected image, and the difference images of the Gaussian filtering under different variances are calculated. The negative part of the difference image is taken as the first background mask, and the positive part is taken as the foreground. The second background mask is obtained by segmenting the corrected image into front and back backgrounds using the Otsu's method; The third background mask is obtained by performing fixed threshold segmentation on the corrected image using the binary method; The actual background mask of the corrected image is calculated by multiplying the difference between the first background mask and the second background mask with the third background mask, and the average brightness of the actual background intensity region of the corrected image is determined. Calculate the ratio coefficient between the fixed background intensity and the average actual background intensity of the corrected image, and stretch the foreground based on the ratio coefficient to obtain an enhanced image.

4. The method of claim 1, wherein the method further comprises: The tilt correction of the region of interest (ROI) in the display panel ripple image includes: After performing binary segmentation of the ROI, the Canny operator is used to calculate the ROI boundary. Set a tilted mask with an angle equal to the diagonal of the ROI; Calculate the intersection of the ROI boundary and the tilted mask to segment out the diagonal boundary; The angles of each side are calculated using the least squares method, and the mean angle is used for tilt correction.

5. The method of claim 4, wherein the method further comprises: The distortion correction of the region of interest (ROI) in the ripple image of the display panel includes: After performing image binary segmentation on the tilt-corrected ROI, the ROI boundary is recalculated using the Canny operator. Set up a horizontal mask and an angled rectangular mask; Calculate the intersection of the ROI boundary with the horizontal mask and the rectangular mask to segment the horizontal and vertical boundaries; Calculate the first-order linear functions of each horizontal and vertical boundary using the least squares method, and calculate the first intersection point between each pair of boundaries; The average of the x-coordinates of any first intersection point and another first intersection point on the horizontal boundary where the first intersection point is located, and the average of the y-coordinates of another first intersection point on the vertical boundary where the first intersection point is located, are taken as the second intersection point of the ROI boundary. Calculate the perspective transformation matrix of the first and second intersection points to perform distortion correction.

6. The method of claim 5, wherein the method further comprises: The rectangular mask is tilted at an angle of 75°.

7. A display panel mura defect detection apparatus, characterized by, include: The correction module is used to perform tilt and distortion correction on the region of interest (ROI) of the display panel ripple image. The image enhancement module is used to enhance the corrected image after tilt correction and distortion correction. The determination module is used to calculate the location information of all defects within the enhanced ROI, and set a periodically changing minimum pixel distance metric for adjacent defects to periodically determine the defects. The determination module calculates the location information of all defects within the enhanced ROI, sets a periodically varying minimum pixel distance metric for adjacent defects, and performs periodic defect determination, including: Determine the x and y coordinates and corresponding numbers of all defects within the enhanced ROI; Calculate the difference between the horizontal and vertical coordinates of adjacent defects, and determine the set A of the indices that satisfy the condition that the difference between the horizontal coordinates is greater than the average width of the defect and the difference between the vertical coordinates is greater than the average height of the defect. Set a periodically changing minimum pixel distance D, and determine the set of local peak indices B where the minimum height difference between the horizontal and vertical coordinates is greater than D; Remove discrete defects belonging to set B from set A.

8. The display panel mura defect detection apparatus according to claim 7, wherein It also includes a calculation module for calculating the internal and external contrast of defects, the calculation module being used for: The enhanced ROI is converted into a first defect binary map BW1; Calculate the center coordinates and inner circle contrast of each defect in BW1; Create a disk-shaped structural element and perform an expansion operation on BW1 to obtain the second defect binary image BW2; Calculate the center coordinates and reference contrast of each defect in BW2; Calculate the distance between the center coordinates of each defect in BW2 and the center coordinates of all defects in BW1 in turn; In BW2, the reference contrast of the defect with the smallest distance is selected as the outer ring contrast of a defect in BW1.

Images