##On-line Learning Gaussian Mixture Models used in Image Simulation##
Represent the intensity distributions of image by a parametric mixture model trained in an on-line (i.e. progressive) manner from a potentially infinite stream of date set. Implement the On-line Learning algorithm used in GMM(Gaussian Mixture Model) based on the image pixel data and find the methods to adjust the model to be more accrate and robust, include using different sampling methods, add SVD constraint and derivate the scalar sigma in updating process.