A multiphase region-based framework for image segmentation based on least square method

We propose a multiphase region-based framework for image segmentation using Least Square Method, by piecewise constant optimal approximations. The basic idea of our model is to build up a minimum error functional by approximating n sub-regions of the original image with n constants respectively. The main contribution of our method is that we introduce weighting matrixes into the region-based model, which can enhance the weight of the specific region while reducing the influence from other regions. Moreover, our method can fast converge, and segment a given image into arbitrary regions under least squares and iterative algorithm. Experimental results show the advantages of our method in terms of accuracy and efficiency in image segmentation.