Wigner based crack detection in textured images

A Wigner based approach is proposed for detecting surface cracks in textural background. The textured region is modeled by the Wigner spectrum. The algorithm consists of three stages: training the system to learn the underlying background textural pattern, obtaining the local variation from the underlying texture and postprocessing to isolate the crack pixels. By this method, crack in random and regular textured background can be detected efficiently. Some natural texture images have been used for evaluating the performance of the algorithm. Promising results are obtained and presented