A LCD Screen Mura Defect Detection Method Based on Machine Vision

Abstract

In the production process, the LCD screen will inevitably have various defects such as dead pixels, had lines, and block defects. Among them, the Mura defect is one of the defects with high misdetection rate because of its low contrast and blurred contour. This paper studies the detection of Mura defect and proposes the following innovations: (1) Compared with other low-pass filters, filtering periodic high frequency texture background by Gabor filters can enhance the local contrast of the Mura region effectively. (2) In order to remove the effect of uneven brightness while retaining the Mura defect, the image is subjected to brightness equalization based on Retinex theory, which enhances the global contrast of the Mura region. (3) Under the premise of retaining defect information, the image is downsampled by selecting appropriate coefficients through experiments, which improves the efficiency of the algorithm. This method was tested with the sample images collected at the industrial site. The detection accuracy reached 93.6% and the average detection time is 0.632 seconds, which meet the requirements of industrial detection. Key Words: Machine Vision, Mura Defect Detection, Gabor Filter, Retinex Theory