Level Set Segmentation Based on Local Gaussian Distribution Fitting

In this paper, we present a novel level set method for image segmentation. The proposed method models the local image intensities by Gaussian distributions with different means and variances. Based on the maximum a posteriori probability (MAP) rule, we define a local Gaussian distribution fitting energy with level set functions and local means and variances as variables. The means and variances of local intensities are considered as spatially varying functions. Therefore, our method is able to deal with intensity inhomogeneity. In addition, our model can be applied to some texture images in which the texture patterns of different regions can be distinguished from the local intensity variance. Our method has been validated for images of various modalities, as well as on 3D data, with promising results.