Texture Defect Detection using Human Vision Perception based Contrast

Abstract

Defect detection in images with periodically repeating patterns is much more complex than that in plain textural images because of high similarity among the repeating patterns. Human vision perception based features are ideal for detecting such defects in such textural images. In this paper, a defect detection method is presented for identifying defects in periodically patterned textures based on Human Vision Perception (HVP) based contrast. Input defective images are split into several periodic blocks and two features, namely, HVP contrast of each periodic block and its absolute difference with the global contrast are used as 2-Dimensional feature space to identify defective blocks using Ward’s hierarchical algorithm. Various real fabric images with defects have been tested using the proposed method and the performance is evaluated in terms of accuracy and computation time. Capability of the proposed method of defect detection is demonstrated by comparison with Gabor wavelet-based method of defect detection with features of neuro-physiological aspects of human vision perception.