Fabric defect detection based on an adaptive bi-orthogonal wavelet

With the development of computer vision technology, computer vision has been increasingly applied to detection and classification of fabric defects automatically. But how to improve the efficiency and reliability of fabric defect detection has been a focal point in fabric inspection research, and it remains challenging. Aiming at edge detection of fabric defects, the designs criteria of an edge detection filter is researched and analyzed, and then an approach of detecting adaptively fabric defect based on bi-orthogonal wavelet is presented by choosing an optimal objective function of maximal average modulus ratio of edge to non-edge on fabric images. The experimental results indicate that the proposed approach has a good ability to detect fabric defect edge and good robustness, compared with the existing methods.