LCD mura detection with multi-image accumulation and multi-resolution background subtraction

We integrated the techniques of multi-image accumulation and multi-resolution background subtraction to detect mura defects in low-contrast and high-noised TFT-LCD images. First, several images of an LCD on a moving product conveyer are contiguously captured and then a synthesized LCD image is used to calibrate the non-uniform illumination of these images. Second, the images are aligned in position to accumulate the gray levels of the pixels which all correspond to a point on the LCD. The multi-image accumulation process enhances the contrast between the mura defects and the background; moreover, visible mura problems due to the view angle and the uneven illumination are also mostly resolved. Third, the multi-resolution backgrounds of the accumulated image are progressively estimated based on the discrete wavelet transform (DWT) and the defect candidates are extracted coarse-to-fine and accumulated by subtracting the background from the accumulated image. The multi-resolution background subtraction strategy speeds the detection process and solves the problem of different sizes of mura defects without reducing the detection rate. Finally, a standard thresholding method is used to "threshold out" the mura defects. The stability and effect of the proposed method are demonstrated in experiments.