An Automatic Detection Method of Mura Defects for Liquid Crystal Display Using Real Gabor Filters

Mura is one types of defects on liquid crystal display (LCD), it is difficult to be detected with regular image processing method. As the existing of respective textural background, the texture analysis approach is helpful. In this paper, the real Gabor filer method and its principles of parameters selecting are studied for the Mura detection. First, in the preprocessing, threshold segmentation, geometric operations and gray interpolation are carried out to pick-up the target normal direction of LCD areas. Then, the real Gabor filters are designed through selecting appropriate filtering scale and orientation, detection results show that the structurally textural backgrounds are removed. Finally, experiments show that the performances of Gabor filters are influenced by its parameters such as central frequencies and mask size. Furthermore, by using real Gabor filters instead of complex, large computational load is saved.