Gpu-based tft-lcd mura defect detection method

A GPU-based TFT-LCD Mura defect detection method, comprising the following steps: (1) establishing, according to original image data, a dual second-order regression diagnosis model based on a studentized residual, and obtaining dual second-order regression background image data; (2) obtaining an influence quantity of each group of data points on a fitted value according to the original image data and the dual second-order regression background image data; (3) removing outliers and influential points from the original image data according to the influence quantity to obtain a new pixel point set; (4) establishing a dual N-order polynomial surface fitting model according to the new pixel point set, and obtaining dual N-order background image data; (5) obtaining a residual image R according to the dual N-order background image data and the original image data, and performing threshold segmentation on the residual image to obtain a threshold segmented image; and (6) performing morphological processing on the threshold segmented image to obtain a corroded and expanded image, thereby effectively segmenting the Mura defect of non-uniform brightness distribution.