Date

Place

[P1-65]
14:00 ~ 15:30
Title:Automatic Micro Defect Detection in Non-Repetitive Patterns for High-Resolution TFT-LCDs
Euiyeol Oh, Dae Keun Kim, Jong Sang Baek (LG Display Co., Ltd., Korea), and Yoonsik Choe (Yonsei Univ., Korea)
Abstract: As TFT-LCD panel becomes larger, micro defects can only be identified at high-resolution and high-magnification (more than 10x) images since various types of micro defects of high-resolution TFT-LCD are not easy to be detected by conventional methods at normal magnification (~7.5x). Previous techniques that use template-matching algorithms are not applicable to highly magnified image of TFT-LCDs since high-magnification pixels (more than 10x) of LCD are non-repetitive. In this study we propose an automatic defect detection system in a fine resolution TFT-LCD image at high-magnification rate. This system is based on two main steps, segmentation and defect detection. A first step is to segment two representative defect types such as pixel-region defect and gate-to-data line (G/D) crossover defect using Wiener Filter. For the defect detection at the pixel-region, we identify defects based on SVD (Singular Value Decomposition) and Local Variance, in that pixel-region is patterned. By choosing the dominant singular values, the reconstructed image is able to remove background patterns and recognize candidate defects. Local variance both amplifies the defects and removes noise.
PDF Download

Type