Infrared Image Segmentation Algorithm Based on Improved Variational Level Set Model

An improved variational level set model is presented for segmentation of infrared images in this paper. Infrared images always have very little priori-information and smooth boundaries or even with discontinuous boundaries, and therefore the segmentation of infrared image is very difficult. We obtain a variational formulation level set model, which can detect contours, both with and without gradient based on global information. The stopping term does not only depend on the boundary gradient of the image, which can overcome effectively the over segmentation. At the same time, this algorithm discards the re-initialization procedure and forces the level set function to be close to a signed distance function, therefore reduces the side effects of re-initialization. The method is used in the segmentation of several kinds of infrared images, and the experiment results reveal the effectiveness of the algorithm listed in this paper.