Blind Non-independent Image Separation Based on Independent Component Analysis
Wu Guo, Peng Zhang, Runsheng Wang · Computational Intelligence and Security · 2007
Blind separation of mixture images which mutually independent has been solved efficiently by some independent component analysis(ICA) methods. But these methods often failed in case of the source images are statistically non-independent. A novel fixed-point FastICA algorithm based on complexity pursuit is presented in this paper and with the algorithm the mixed images which not mutually independent can be separated successfully. Experimental results demonstrate the efficiency of our proposed method.