基于区域灰度统计信号处理图像融合方法_王睿.pdfVIP

2010 2 F ebru ary 20 10 36 2 Journal of Be ij ing U n iversity o f A eronautics and A stronautics V o l. 36 N o12 王 睿 王 林 袁 艳 ( , 100 19 1) 针对多源图像融合问题, 提出了一种在多分辨率框架下基于区域 灰度特 征统计信号的融合算法. 利用图像灰度特征的区域生长法对源图像进行区域分割, 并以裂缝边 缘作为特征区域的闭合边界, 对源图像与分割结果的区域映射图作多分辨率变换. 在图像低频 部分, 以联合区域映射图为指导, 在区域 建立信号与噪声的高斯混合分布模型, 利用期望极 大化( EM, Expectat ion M ax mi izat ion) 算法迭代估计噪声模型分布参数, 获得低频融合结果; 在 图像高频部分, 根据系数在区域映射图上的位置差异分别采用窗口系数加权平均法和系数绝 对值选大法进行融合, 将低频和高频融合结果反变换得到最终融合图像. 融合结果表明 该方 法是可行和高效的, 且比其他图像融合方法具有更好的性能. 图像融合; 区域生长; 裂缝边缘; 期望极大化; 多分辨率框架 TP 391. 41 A 1001-5965 ( 2010) 02-0140-05 Reg ion-based statistica l s igna l process ing scheme fo r mi age fus ion W ang Ru i W ang L in Y uan Y an ( S chool of In strum ent S cien ce and O pto-electron ics Eng ineering, Beij ing U n iv ersity of A eron aut ics and A stronau tics, B eijing 100 191, Ch ina) A bstract A new mi age fu sion schem e b ased on reg ion stat istical signal processing w as proposed. The re- g ion grow ing technique using gray- level clu stering w as em ployed to segm ent the source mi ages into different re- g ion sw hose borderline represented w ith crack edge. T he reg istered source mi ages and their segm ented m ap- p ing w ere decom posed into a mu lt-i resolution represen tation w ith both low-frequ ency coarse inform ation and h igh-frequ ency de tail inform ation resp ect ive ly. T he exp ectation m ax mi izat ion algorithm m ode led w ith no ise sta- tist ic d istribu tion w as u sed to fu se the low-frequ ency coarse inform at ion of the reg istered mi ages, w h ile the m atch and sa lience m e