An Active Contour Model Based on Local Entropy for Image Segmentation with High Noise

Active contour models (ACMs) have been employed extensively in the area of image segmentation. Howbeit, ACMs exist some disadvantages including slow evolution, sensitivity to intensity inhomogeneity and noise. Therefore, an ACM based on local entropy is put forward to segment images with high noise and inhomogeneous intensity. Specifically, the local entropy fitting image is firstly introduced to constrict different noise kinds and levels when preserving image detail information. The bias correction energy formulation is constructed through employing the local entropy fitting image to estimate bias field for better correcting the massive inhomogeneous intensity distribution. Finally, an enhanced regularization term and the average filtering are applied to eliminate instability in numerical calculations during the evolution of level set function. The comparative experiments conducted on synthetic and real images with high noise and intensity heterogeneity indicate the better accuracy and robustness of the introduced model.