基于方向导数的多光谱图像快速融合新算法  

A Novel Fast Multispectral Image Fusion Algorithm Based on Directional Derivative

作　　者：陈武[1] 靳海兵[1] 吴政[1] 王祥涛[1] 

机构地区：[1]西北工业大学电子信息学院,陕西西安710072

出　　处：《计算机仿真》2009年第10期257-260,共4页Computer Simulation

摘　　要：针对提升小波变换的特点,研究图像融合规则,提出了一种新的基于方向导数的多分辨多光谱图像快速融合算法。首先,利用提升小波变换得到待融合源图像的多分辨分析;然后以小波域方向导数为判据,在图像的多分辨分析的相应各级上进行融合,得到融合图像的多分辨分析;最后,利用提升小波反变换重构融合图像。使用多光谱图像对算法进行验证并与方向对比度算法进行对比,结果表明,融合图像完好的融合了源图像的信息,且算法运算时间短。According to the characteristics of lifting wavelet transform, the image fusion rules are researched, and a novel fast image fusion algorithm based on Directional Derivative（DD） for Muhispectral images is proposed. The implementation of the method includes three steps ： First, the input multispectral images are decomposed by lift- ing wavelet transform, meanwhile, the multiresol,.,tion sequences of each input imges can be obtained. Secondly, ac- cording to the directional derivative criterion, the multiresolution analysis of the fused image can be obtained at corresponding levels of the multiresolution analysis of input images. Finally, the output image can be obtained through inverse lifting wavelet transform. This algorithm is tested and compared with directional contrast by fusing the multispectral images from same scence. The experiment shows that the fused image can preserve the details of each input image successfully.

关 键 词：图像融合 提升小波变换 方向导数 方向对比度 

分 类 号：TN911[电子电信—通信与信息系统]