The region based MMTD energy function for image segmentation

This paper presents a region-based model based on measure of medium truth degree for image segmentation. Firstly, a new energy function based on measure of medium truth degree is constructed. To enhance the robustness against noise, a noise penalty term which is built by spatial distance measure is embedded to the conventional active contour energy function. Then local information is added to the internal and external energy term of the conventional active contour energy function to deal with intensity inhomogeneity images. Finally, to obtain more accurate and smoother boundary, a new stop function is introduced into the boundary smooth term of the conventional active contour energy function. Experiments results demonstrate that relatively complete and accurate boundaries can be obtained by the proposed model compared with the state-of-art methods on aurora images, images with intensity inhomogeneity, images with multi-objects, natural images, medical images.