Mura Defect Of TFT-LCD Detection Method Based On Spatial Standard Observer

In the industrial display panel inspection,Mura defect detection of a thin film crystal liquid crystal display(TFT-LCD)is a key point.Mura defect is not only low contrast and edge blur,but also uneven brightness,often contains periodic textures,these features make it more difficult to detect,nowadays commonly used detection methods of machine vision,although the method detection effect is good,but only Pay attention to the accuracy of detection,and ignore the habit of human eye vision.In this paper,a Mura defect detection method based on the spatial standard observer's TFT-LCD is proposed for the characteristics of Mura defects,the requirements of industrial inspection,and the visual perception of the human eye.This method is mainly applied to human eye contrast.Sensitive function performs image filtering to produce a minimally detectable difference map that is in line with human visual perception,and then enhances the minimum detectable difference map.Finally,the image segmentation technique is used to segment the defect.Experiments have shown that this method can not only effectively detect where the Mura defect is,but also suitable for human visual perception.The main work of the manuscript is as follows:(1)Image down-sampling.In the actual industrial inspection,in order to ensure the detection accuracy of the panel,a high-precision industrial camera is generally used and the bitmap format is used.Therefore,the images collected by the camera are generally very large,approximately tens of megabytes in size,in order to ensure detection efficiency and to reduce the performance requirements of the device,the image can be down-sampled in advance so that the image data can be reduced.In order to select a suitable sampling ratio,a variation function is used to count the period of the image grayscale,and the calculated vertical and horizontal grayscale periods are taken as the mean sampling ratio,and then the mean down-sampling is performed.This method can not only compress the image size.Reduce the amount of calculation,but also to ensure a certain image quality.(2)Texture inhibition.When a high-precision camera is used,the acquired image is enlarged and a lattice-like periodic texture is found,or the captured image itself is textured.These textures cover out some weak Mura and cause missed detection.In this paper,the frequency spectrum filtering method is used to analyze the frequency spectrum,and the energy function of the filter radius is constructed.The critical radius is found by analyzing the curve of the energy function along with the filter radius,and finally the frequency domain filtered image is inverse transformed to the spatial domain,resulting in a texture-suppressed image.(3)Research and improvement of space standard observers.Space standard observer This method involves physics,psychology,human vision,etc.It is a system with high gold content.This paper studies the features of human-eye vision system,the correction of image brightness and contrast,the establishment of human-eye contrast model,and the minimum detectable difference threshold.Based on this method,the minimum detectable difference map is improved.Enhancements were made to highlight Mura defects and a simple Mura defect criterion was proposed.(4)Mura defect segmentation.Image segmentation is the final step of the design,and there are many methods,mainly watershed method,morphology method,cluster analysis method,Otsu method,etc.This article takes a different approach,aiming at the monotonous background and minimal awareness of the acquired image itself.Due to the uniform gray feature of the difference image,a method of image segmentation based on the average value and standard variance of the gray image is proposed.In the generated image,the mean value represents an overall uniform background,and several times the standard deviation represents the defect area with different backgrounds,so only an appropriate threshold value needs to be selected for efficient segmentation.At the end of the manuscript,according to the method designed in this paper,Matlab's Mura's defect simulation experiment was performed to verify the method.The results show that this design method is very effective and can effectively detect Mura defect images that contain noise,repetitive textures,uneven brightness,and other interference factors.It can not only accurately identify defects,but also conform to the eyes' visual perception habits.The results of the machine test are consistent with the results of the human eye test,and the corresponding Mura defect evaluation criteria are given.