动态WNMF和在图像融合中的应用研究.pdfVIP

23 9 Vo.l 23 No. 9 2010 9 CH INESE JOURNAL OF SEN SORS AND ACTUATORS Sept. 2010 * DynamicWeightedNon-NegativeMatrix Factorization and ItsUsingResearch in ImageFusion * LIUShaopeng, HAO Qun , SONG Yong (Optoelectronic College, Bejiing Institute of echnology, Bejiing 100081, China) Abstract: The mi age fusion algorithm based on standard non-negativematrix factorization has lmi ited global feature ex- traction ability, resulting in the low contrast and poor visual effect of the fusion result. In order to mi prove the fusion effect of non-negative matrix factorization, a novel mi age fusion algorithm for infrared and visible mi ages based on dy- nam icw eighted non-negativematrix factorization is proposed. Thew eighted coefficients are designed for emphasizing the mi portant characteristics of the source mi ages, and they aremod ified after every iterative accord ing to the relative mi por- tance variation of different areas, so the feature extraction ability of weighted non-negative matrix factorization is en- hanced obviously. Comparedw ith standard non-negativematrix factorization based fusion algorithm, proposed algorithm mi proves the visual effect, and the average gradient of the fusion resu lt mi provesmore than 36% and the standard devia- tion mi proves more than 17%. eywords: mi age fusion; feature extraction; dynam ic weighted non-negativematrix factorization; mutation degree EEACC: 6140C doi: 10. 3969/j. issn. 1004- 1699. 2010. 09. 012 * WNMF * 刘少鹏, 郝 群 , 宋 勇