Application of human visual system models to LCD image analysis

Abstract

Automated analysis of the segment LCD images may be utilized to detect hardware or software related problems of the device or LCD itself. The malfunction may be revealed through detection of faulty segments. Difficulties to automate image analysis arise from existing intensity and geometric distortions. We adopt concepts of biological visual systems to improve performance of computational system. In this paper we compare conventional and bio-inspired local feature detectors along with descriptors, and introduce a new feature matching approach based on evidence accumulation. Finally we construct specialized algorithm for alignment of LCD images. The performance of the proposed algorithm has been evaluated on the set of semi-synthetic images. The comparison with common image registration techniques shows elimination of image misalignments.