Robust defect detection method for a non-periodic TFT-LCD pad area

Abstract

In this work, we propose an effective inline inspection method for non-repeating patterns on TFT-LCDs which can be quickly applied to current manufacturing processes and easily automated. Given that the TFT-LCD active area is composed of regularly spaced repeating patterns (pixels) inline inspections are possible using relatively simple methods. However, the pad area, which transfers digital signals to the TFT active pattern, is composed of patterns with complex shapes in the form of either shape changing patterns or repeating patterns with varying periods. This makes it very difficult to examine complex patterns in the pad area, because repeating patterns have to be divided into areas with similar repeating frequencies and the inspection parameters have to be set differently, and can cause a bottleneck preventing inspection yield improvements. In this paper, an inspection method using a discrete Fourier transform (DFT) filter and a local threshold binarization method is proposed. We show that the proposed method works on an image having both impedance matching and fan-out areas at a varying frequency with a single parameter set. This setting of parameters also can be done in an automated way.