Defect detection in textured images

Abstract

Texture defect detection is an important research area which can find direct applications in industry. The research work reported in this thesis includes a study of the texture defect detection problem. The study looked into the nature of defect and its detection requirements. Requirements on features that are suitable for defect detection are discussed. Feature stability is believed to be more important in defect detection than in texture segmentation and classification. This motivates the concept of using a local interaction neighbourhood that enables stable feature extraction.