Active contour model based on local bias field estimation for image segmentation

Abstract The active contour model is a commonly used image segmentation method. When applied to complex images, such as images with grayscale inhomogeneity, most existing active contour models produce poor segmentation results. In order to solve this problem, we propose an active contour model based on local bias field estimation (LBFE), which makes the improved model better able to segment complex images. Firstly, we propose a new function to compute the value of bias field with fuzzy c-means clustering algorithm. This computation is completed before the iteration, which greatly improves the compute speed. Secondly, compute minimization with the energy function in the bias correction model (BC) proposed by Li et al. Thirdly, a new variational level set function is proposed to limit the segmentation range and greatly improve the robustness. Experiment results have proved that the proposed model not only segments images with intensity inhomogeneity effectively and shortens time spent, but also shows a better robustness to initialization and a higher segmentation accuracy than other classic models.