Defect detection on semiconductor wafers using image processing techniques

Abstract

Defect detection in semiconductor wafers is an important step in the semiconductor manufacturing process. Since manual defect detection takes much time and is less accurate, various image processing techniques for defect detection have been developed. In this thesis, we implement two defect detection methods. The first method is based on wavelet transform. It calculates the wavelet transform modulus sum (WTMS) for suspicious pixels for detection. The second method is based on self-similarity. The method generates a residual image from a source image by self-similarity based background subtraction, and defects are determined through a statistical method. We apply these methods to several test images and discuss their pros and cons.