APPLYING DISCRETE COSINE TRANSFORM AND GREY RELATIONAL ANALYSIS TO SURFACE DEFECT DETECTION OF LEDS

Abstract

ABSTRACT Light-emitting diodes (LEDs), owing to their lower power requirements, higher efficiency, and longer lifetimes, are widely used in modern electronic devices. Nevertheless, tiny defects that often appear in the surface of LEDs impair not only their appearances but also their functions. This paper proposes a global approach for the automated visual inspection of tiny surface defects in non-diffused LED encapsulations. The proposed method, taking advantage of the discrete cosine transform (DCT) and grey relational analysis techniques, overcomes the difficulties of inspecting tiny defects on uneven illumination images of LEDs. We apply the grey relational analysis to the frequency components in the DCT domain, and select the large-magnitude frequency components that represent the background texture of the surface. Then, by reconstructing the frequency matrix without the selected frequency values, we eliminate not only random texture but also uneven illumination patterns and retain anomalies in the restored image. Experimental results demonstrate the effectiveness of the proposed method in inspecting tiny defects in non-diffused LED surfaces.