A convex active contour model driven by local entropy energy with applications to infrared ship target segmentation

Abstract

To segment ship target in infrared (IR) images, a convex active contour model based on local image entropy is proposed in this paper. Firstly, local image entropy based on kernel function is obtained; on this basis, the convex energy functional is built using the variational level set method in the convex set. Because of the introduction of entropy, the proposed method can protect the image edge and enhance the ability to deal images with heterogeneity. At the same time, the convex energy functional can get a global minimum and has robustness against initial curve placement. Compared with the state-of-the-art methods, experiment results demonstrate the performance and effectiveness of the proposed method.