基于模糊推理的小波域图像融合规则  

Wavelet Domain Image Fusion Rule Based on Fuzzy Reasoning

作　　者：吴佳鹏[1] 杨兆选[1] 苏育挺[1] 王曾敏[1] 陈杨[1] 

机构地区：[1]天津大学电子信息工程学院,天津300072

出　　处：《天津大学学报》2007年第12期1416-1420,共5页Journal of Tianjin University(Science and Technology)

基　　金：天津市自然科学基金重点资助项目(07JCZDJC05800)

摘　　要：现有的2种主要的小波域融合规则,即选择绝对值最大系数规则和选择与加权平均相结合的规则,都存在着融合信息不完整和易受噪声干扰等缺点.为克服这些缺点,使融合图像包含尽可能多的源图信息,提出了基于模糊推理的小波域融合规则.此规则根据源图像小波系数的局部区域特征,通过模糊推理确定各源图像相应系数的权值,再对相应系数加权平均得到融合后的小波系数.实验结果表明,使用基于模糊推理的融合规则融合的图像,在峰值信噪比以及与源图像之间的互信息量等方面均优于使用2种现有规则进行融合所得的图像.新规则有效地克服了现有规则的缺点,具有更好的信息完备性和鲁棒性,尤其在源图像受到噪声污染的情况下,其优势更加明显.Two main wavelet domain fusion rules at present, which are selection of the absolute value maximum and combination of selecting and weighted averaging, are sensitive to noise and have incomplete information of fused image. To overcome the drawbacks to extend as much information of source images as possible into fused image, a wave domain image fusion rule based on fuzzy reasoning was proposed. According to local area features of wavelet coefficients in source images, the weights of corresponding coefficients were obtained through fuzzy reasoning, then the corresponding coefficients were averaged to obtain the fused wavelet coefficient according to the weights. Experimental results demonstrate that the mutual information and peak signal to noise ratio （PSNR） of the image fused with the rule based on fuzzy reasoning are better than those of the images fused with the two main rules at present. The new rule effectively overcome the drawbacks of the two present rules, and have better information completeness and robustness, especially in case that source images are stained by noise.

关 键 词：图像融合 小波变换 模糊推理 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]