Multiscale Detection of Defect in Thin Film Transistor Liquid Crystal Display Panel

Abstract

A new automated inspection algorithm is proposed for detecting blob-Mura defects based on multiscale in a thin film transistor liquid crystal display (TFT-LCD) panel. As such, new kernels with different sizes are defined and then used to detect blob-Mura defects with varying sizes and brightness levels. To extract a seed point, an adaptive multilevel-threshold method is employed. Initially, smaller kernels are used to detect smaller defects, then gradually larger kernels are applied to detect larger defects. Through simulation it was verified that the proposed algorithm has a superior capability for detecting blob-Mura defects.

Defect detection in periodic patterns using a multi-band-pass filter

“…The masks for the designed multi-band-pass filter are located at the origin, fundamental frequency coordinates, and the coordinates of integer multiples of the fundamental frequency. The golden template of a possibly defective periodic pattern can be generated from the multi-band-pass filter given in (14). In theory, the multi-band-pass filter H (u, v) can be implemented either using the DFT or its convolution mask.…”

“…In theory, the multi-band-pass filter H (u, v) can be implemented either using the DFT or its convolution mask. Once the golden template is generated using (14), the image with the defects can easily be obtained by comparing the test image against the golden template. Unlike low-pass filters, the present approach can sharply discriminate the energy distributions of the defect pattern and the periodic pattern in the high frequency range so that both images will be sharp and accurate.…”

“…Most existing approaches for defect inspection in periodic patterns can be divided into three main classes: (1) template matching [1][2][3][4][5][6][7][8][9][10][11][12][13]; (2) neural networks [14][15][16][17][18][19][20]; and (3) filter approaches [21][22][23][24][25][26][27][28][29][30][31][32][33]. A few other studies have reported approaches other than the abovementioned [34].…”

“…In approaches using neural networks and statistical methods [14][15][16][17][18][19][20], it is difficult to discriminate the defective area with periodic backgrounds. Therefore, neural networks have limited applications in the mura inspection of thin film transistor-liquid crystal displays (TFT-LCDs) [14,15] and in the relatively simple patterns of light emitting diodes (LEDs) [16,17].…”

“…Therefore, neural networks have limited applications in the mura inspection of thin film transistor-liquid crystal displays (TFT-LCDs) [14,15] and in the relatively simple patterns of light emitting diodes (LEDs) [16,17]. In particular, the support vector data description (SVDD) method using a fuzzy penalty function has been proposed to inspect defects in gate electrodes [15,18].…”