文摘
We rewrite the cost function and derive a novel updating of level set function based on probabilistic principles. We propose two novel geometric priors from active contours, and both of them have advantages. We choose the suitable prior as needed to obtain level set function in EM algorithm, which reduce the computational cost. Updating of level set functions and estimation of statistical model parameters are run alternately. We use the student's-t mixture model with heavy tail to enhance the robustness.