A Level Set Method Combined with Gaussian Mixture Model for Image Segmentation

摘要

Chan-Vese (CV) model promotes the evolution of level set curve based on the gray distribution inside and outside the curve. It has a better segmentation effect on images with intensity homogeneity and obvious contrast. However, when the gray distribution of image is uneven, the evolution speed of the curve will be significantly slower, and the curve will be guided to the wrong segmentation result. To solve this problem, a method to improve CV model by using of Gaussian mixture model (GMM) is proposed. We use the parameters of the Gaussian submodels to correct the mean value of grayscale inside and outside the curve in the energy function. The target region can be quickly segmented in the images with complex background gray distribution. Experimental results show that the proposed algorithm can significantly reduce the number of iterations and enhance the robustness to noise. The level set curve can quickly evolve into target region in the images with intensity inhomogeneity.