一种塔形分解的红外与可见光图像融合方法  

Fusion on Weighted Median Pyramidal Decompositions

作　　者：刘卫光[1] 周利华[1] 

机构地区：[1]西安电子科技大学多媒体研究所,陕西西安710071

出　　处：《红外技术》2004年第4期31-35,共5页Infrared Technology

基　　金：十五国防科技(电子)预研项目资助(编号:413160501)

摘　　要：针对红外图像对比度低、背景噪声较大的情况,提出了一种基于加权中值塔形分解的红外与可见光图像融合算法,同传统拉普拉斯塔形分解方法相似对图像的分解分成四步:中值滤波、降采样、插值放大、带通滤波,形成多尺度、多分辨率的塔形近似图像序列和细节图像序列。两种传感器图像经过中值塔分解的图像序列,通过基于局部能量选择的融合规则和融合算子形成融合图像塔,然后进行中值塔形图像重构,完成图像融合。与已有塔形分解图像融合方法相比,不仅保留塔形分解的优点,融合图像精确,而且对脉冲型噪声污染的传感器图像具有较强的抑制作用。实验验证,这种融合方法在图像信噪比较低的情况下,也能得到较好的融合效果。A novel fusion algorithm based on weighted median pyramidal decompositions of multisensor image is presented,similar to Laplacian pyramid proposed by Burt and Adelson,the decomposition contains median filtering、downsampling、interpolation、band filtering. Then,the approximation and the detail image pyramid of multiscale and multiresolution are obtained. Multisensor images are decomposed three levels,at which pyramid of image fusion is gotten though the fusion rule of local analysis window energy. Image fusion is accomplished by the median pyramidal reconstruction. It has the advantages of pyramid method, overcomes the drawback of traditional Laplacian pyramid method and is sensitive to salt and pepper noise. Finally,the results of fusion experiments by IR image and visible image registered perfectly are satisfied.

关 键 词：加权中值塔形分解 图像融合 中值滤波 脉冲噪声 传感器 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]