Skeletonize multi width ribbon-like shapes based on difference images and frenet frame

In this paper, we propose a novel skeletonize method for multi width ribbon-like shapes. The two difference images with vertical and horizontal directions respectively are computed to locate the singularities of the original image. Then the gradient directions of these singularity points are computed. The skeletonization can be obtained by finding the centers of via pair wise singularity points using their gradient directions. The most attracting nature of this method is that it can handle multi width shapes, which is very difficult in conventional methods. The experimental results show that our method is a ldquogoodrdquo method closing to human conception.