The application of classical gaussian mixture model to image segmentation has highly computer complexiton and have not taking into account spatial information except intensity values. A image segmentation based on Gaussian mixture model with sampling and spatially information is proposed in order to solve this problem. First, a spatial information function is defined as the neighbour information weighted class probabilities of very pixels; Secondly, the sampling theorem is given in this paper,and the size of the minimum sample has been derived according to the smallest cluster and cluster number; Finally, image pixels are sampled based on the size of the minimum sample to estimate the parameter of model , which are classifed to different clusters according to bayesian rules. The experimental results show the effectiveness of the algorithm.
Variational Bayesian Inference for Finite Inverted Dirichlet Mixture Model and Its Application to Object Detection