Research on detection of fabric defects based on singular value decomposition

Singular value decomposition technique is widely employed in feature analysis due to its strong capability of feature expression. Aiming at detection of fabric defects, this paper gives an approach for the fabric defects extraction in an image based on the theories of singular value decomposition, and proposes the corresponding algorithm. Firstly, singular value decomposition is performed on sub-image of the entire image, size of a rectangle window so that the average of singular values of every sub-image is obtained. Then, according to last step the average of singular values of all of sub-image is calculated. Finally the fabric image is segmented by means of a threshold related to the average of singular values and the defects could be detected. By using singular value decomposition, the complexes of operation are reduced, and noise issues of the image may be overcome. Validity and feasibility of this approach is proved through several experiments of fabric defects detection.