基于区域特性的Contourlet域多聚焦图像融合算法.PDFVIP

41 4 西 安 交 通 大 学 学 报 Vol. 41 l 4 2007 4 JOU RNAL OF XIcAN JIA OTONG U NIVERSITY A pr. 2007 Contourlet 镠, 郭宝龙, 倪伟 ( , 710071, ) : 利用 Co tourlet 变换的多尺度局部化方向性和各向异性等优点, 提出了一 基于区域特 性的Co tourlet 域多聚焦图像融合算法. 该算法将源图像分解至Co tourlet 变换域, 在不同尺度 不同方向的子带中结合区域特性进行图像融合, 低频和高频子带中分别采用区域方差和区域能量 作为融合规则, 最后通过反变换得到融合图像. 实验结果表明, 所提出的算法能够更好地提取原始 图像特征, 融合后的图像具有更好的主观视觉效果, 与经典的梯度金字塔算法和小波变换算法相 比, 新算法的均方误差最大值仅为前二者的341 8%和421 6%. : 图像融合; Co tourlet 变换; 多聚焦; 区域能量 : TP941 : A : 0253O987X( 2007) 04O0448O05 Multifocus Image Fusion Algorithm Based on Region Statistics in Contourlet Domain Ya g Liu, Guo Baolo g, Ni Wei (I stitute of I tellig e t Co trol a d Im age E gi eeri g, Xidia U iversity, Xic a 710071, Chi a) Abstract: Utilizi g the co tourlet. s adva tages of multiscale, localizatio , directio ality a d a i- sotropy, a multifocus image fusio algorithm based o regio statistics i co tourlet domai is de- veloped. Source images are firstly decomposed to the domai of the co tourlet tra sform, the im- age fusio is the impleme ted i subba ds w ith differe t scale a d directio combi i g w ith re- gio statistics. Regio al varia ce a d local e ergy are adopted as fusio rules i low pass a d high- pass subba ds, respectively. Fi ally the fused image is obtai ed through i verse tra sform. Ex- perime tal results show that the proposed method ca detect image features more effectively a d the fused image has better subjective visual effect. Compared to traditio al gradie t pyramid algo- rithm a d w avelet based algorithm, the maximum of mea squar