A variational model of multiphase segmentation for images with Gaussian noises and its Split Bregman algorithm

In this paper, we propose a variational multiphase segmentation model for images with Gaussian noises. Its data term for parameter estimation is based on density functions of Gaussian distribution and its length term of active contours for n regions division is based on n binary labeling functions. We design the Split Bregman algorithm for the sub-problem of minimization on every labeling function based on the convexified version, and obtain the final solution via soft thresholding technique. Numerical examples validate the model and its algorithm finally.