Image segmentation for defect detection based on level set active contour combined with saliency map

In this paper we propose a new image segmentation method to detect surface defects with small shapes in images characterized by low contrast level and affected by non-uniform illumination. The proposed segmentation method is based on level set active contour approach where a center-surround feature saliency map used to suppress the clutter background and highlighting the potential objects is incorporated in the level set active contour formulation. Experimental results on different industrial images like welding and steel demonstrate the robustness and the accuracy in defect segmentation comparing with others active contour models.