Automatic TFT-LCD mura defect inspection using discrete cosine transform-based background filtering and ‘just noticeable difference’ quantification strategies

Abstract

An innovative mura defect detection methodology for a thin-film transistor liquid crystal display (TFT-LCD) is developed for automatic inspection of mura defects using the discrete cosine transform (DCT) principle and background image filtering strategy. Efficient and accurate surface defect detection on flat panel display (FPD) panels has never been so important in achieving a high yield rate of FPD manufacturing. Detecting blob-mura defects in an LCD panel can be difficult due to non-uniform brightness background and slightly different brightness levels between the defect region and the background. To overcome this problem, a DCT-based background reconstruction algorithm was developed to establish the background image separated from the defects. The significant level of mura defects can be rationally quantified using the just noticeable difference (JND) definition. Actual performance of the developed method was evaluated on industrial LCD panels containing natural mura defects. Results of experimental tests verified that the proposed algorithm has a superior capability for detecting mura defects efficiently and accurately.