Blind Source Separation for Remote Sensing Images based on the Improved ICA Algorithm
Di Shen, Chengfan Li, Jingyuan Yin, Junjuan Zhao, Dan Xue · 2015
In consideration of some problems including the independence and invariance of components, no noise assumption and the uncertainty of the final solution as well as the inconsistency of the features of remote sensing data in traditional independent component analysis (ICA) model, we put forward a blind source separation algorithm for remote sensing images using variational Bayesian ICA.In the proposed method, the Bayesian network is introduced into the ICA model, the Bayesian inference is used to complete the study of unknown hidden variables, and the computation is optimized by combination with the variational approximation method.Finally, the proposed method is validated by simulation and real remote sensing image tests.The result shows that the variational Bayesian ICA algorithm has both good stability and separation effect, and it overcomes the deficiency of the traditional ICA method in remote sensing application.