Research of Image Separation Based on Improved Independent Component Analysis
Si-zu Hou, Ming Wang, Xincheng Tian, Xu Wang, Liyan Huang · 2006
The basic theory of ICA and algorithm of FastiCA are briefly introduced. A variant of Newton's iteration method is given. It is shown that the order of convergence of the new method is three. A new algorithm combined Newton's iteration method is proposed to increase the convergence speed. The image separation simulation shows that the iterations of improved algorithm reduced 14.6 percent with the correspondent separation performance