Research On Mura Defect Detection Technology Of TFT-LCD Panel Based On Background Reconstruction

In recent years,the Internet of Things has flooded into various fields with lightning speed.As an important medium for human-computer interaction,liquid crystal display(TFT-LCD)is almost everywhere,and its display quality is increasingly being valued by various manufacturers.Mura is a special defect in TFT-LCD,which is characterized by cloud-like,low contrast and blurred edges.In view of this defect,the paper proposed a detection algorithm based on background reconstruction.The main research contents are as follows:(1)Mura defect image preprocessing.Firstly,several methods of image denoising are compared by experimental analysis.Then,in order to make our algorithm detection results more in line with human visual perception,the concept of spatial frequency and human contrast sensitivity function is introduced,and the contrast sensitivity function is used for image denoising.(2)Background reconstruction algorithm design.In this paper,by studying discrete cosine transform method,singular value decomposition method and polynomial surface fitting method,a high precision background fitting method based on Mura defect preremoval is proposed.Firstly,the discrete cosine transform method and the singular value decomposition method are used to roughly divide the Mura defect,and then logic and operation is performed on the two divided images,and the obtained result is used as a pre-removal region.Then,using all the data points except the pre-removed area for surface fitting,a high-precision background image is generated.In order to optimize the running efficiency of the algorithm,based on the minimum number of points required for polynomial fitting,the data points are equally sampled,and the background image is reconstructed using the sampling points,which shortens the time-consuming of the algorithm and its accuracy remains almost unchanged.(3)Contrast enhancement and image segmentation.The contour of the image after subtracting the original Mura image by the background image is still blurred.At this time,based on the comparison of several image enhancement methods,we propose a dual gamma piecewise exponential transformation method based on Otsu,which suppresses the background information of the image and meanwhile the Mura area has been enhanced.Based on the experiment,we selected the optimal range of gamma parameters.Image segmentation is the last step of the proposed algorithm.Here we study two most classic image segmentation methods: maximum inter-class variance method and fuzzy C-means clustering method.Combining the advantages of the two algorithms,an improved image segmentation algorithm is proposed.The threshold obtained by the Otsu method is used as a reference point to set the initial cluster center,and then continually corrects it by iteration,which not only accelerates the convergence speed of the algorithm,but also achieves a better segmentation effect.Finally,three representatives of Mura images,namely point Mura,line Mura and ring Mura,are used as experimental samples,and tested according to the designed algorithm flow.The quantitative results show that the Mura defect detection algorithm based on background reconstruction has high precision and robustness.