Image analysis is used to detect and measure targets in images in order to get impersonal information and establishes the description to images and objects. Non-negative Matrix Factorization(NMF),as a new method of robust characterization analysis,has wide applications in many fields.By studying NMF's basic principle,a new method,Fixed Point NMF(FPNMF),was proposed and general methods of NMF were exlarged.Many algorithom rules using different divergneces were studied.In addition, some difficult problems were also discussed.Finally,images got in special conditions were analysed and pratical results were found.Fixed Point NMF features identifications of the real images is implemented by the observed images directly and also as in the Blind Source Separation.By processing the natural gray images in the simulation,the results show the efficiency and robustness of this approach we made. |