Abstract

This research proposes a new approach that applies discrete cosine transform (DCT) based enhancement for the detection of micro flaws on electronic chips. A two-stage decomposition procedure is proposed to extract an odd-odd frequency matrix after a digital image has been transformed to DCT domain. The cumulative sum algorithm is then applied to detect the transition points of the gentle curves plotted from the odd-odd frequency matrix. After the transition points are determined, the proper radius of the cutting sector is computed and the high-pass filtering operation can be performed. The filtered image is then transformed back to the spatial domain. Thus, this study effectively attenuates the global random texture pattern and accentuates only micro flaws in the restored image. Finally, the restored image is segmented by an entropy method and some features of the detected flaws are extracted. Experimental results show that the proposed method achieves a high 96.24% probability of correctly discriminating micro flaws from normal regions and a low 0.15% probability of erroneously detecting normal regions as defects on random textured surfaces of electronic chips.

Key words: Industrial engineering, surface micro flaw inspection, electronic chips, quality control, machine vision.