Decomposition of Mixed Pixels Based on Bayesian Self-Organizing Map and Gaussian Mixture Model

A new method for the decomposition of mixed pixels of mult-i channel remote sensing images is proposed. The proposed method introduces the Bayesian self-organizing map into the problem of the decomposition of mixed pixels. It estimates the Gaussian parameters by minimizing the Kullback-Leibler information metric,and carries out the unmixing with the Gaussian mixture model.In order to obtain high unmixing precision,the effective range of Gaussian distributions needs to be extended,and the 3σ variances adjustment method is adopted to solve the problem. In addition,the used unmixing model satisfies two constraints,abundances non-negative constraint and abundances summed-to-one constraint,which are demanded in automatical decomposition of mixed pixels. Experimental results show that the proposed method is able to obtain good unmixing results,and robust to noises.