Variational Model for Image Segmentation

Image segmentation is a fundamental problem in the field of image processing and computer vision with numerous applications. In the recent years, mathematical models based on partial differential equations and variational methods have led to superior results in computer vision. In this paper, we present a variational model which uses total variation as a regularization for image segmentation, and develop a primal-dual hybrid gradient algorithm to our model. We discuss images in the cases of Gaussian, Poisson and multiplicative speckle noise, the performance of our model is illustrated by experimental results on synthetic and real data. The proposed model can also preserve small anatomical structures better than Alex Sawatzky's region-based segmentation model with less CPU times.