Active contour model based on local and global Gaussian fitting energy for medical image segmentation

Abstract An improved active contour model is proposed for medical image segmentation in this paper, which integrates the local and global intensity information of the image effectively, with the local fitting term, the active model can attract the contour to stop at the true image edges, the global fitting term is based on the statistical numerical function and level set method. by subtracting the Gaussian convolution image with the original image, the difference images is used to replace the original image in the evolution equation, and the average intensity of the difference image inside and outside the contour is also used to substitute the average intensity of the original image during the level set evolutionary process, the experiments shows the proposed method can have a better segmentation performance with less iterations while dealing with the medical images with intensity inhomogeneity.