非抽样轮廓波变换域图像融合方法研究  

Image fusion based on nonsubsampled contourlet

作　　者：张雄美[1] 李俊山[1] 易昭湘[1] 张铁军 

机构地区：[1]第二炮兵工程学院,西安710025 [2]驻上海地区军事代表室,上海200233

出　　处：《电光与控制》2008年第7期16-19,共4页Electronics Optics & Control

基　　金：国家自然科学基金资助(60572080)

摘　　要：轮廓波变换由于不具备平移不变性,应用于图像融合会产生褶皱现象。针对轮廓波变换的这一不足,提出了一种基于非抽样轮廓波变换的图像融合算法。首先由非抽样金字塔和非抽样方向滤波器组对源图像进行多尺度、多方向分解,然后对得到的变换系数进行融合处理得到融合系数,最后对融合系数进行重构即可得到融合图像。算法将具备平移不变性的非抽样轮廓波变换应用于图像融合,消除了褶皱现象,获得了更好的融合效果。实验结果表明,与其他基于多分辨率分解的融合算法相比,该算法显著减小了融合图像的MSE及ΔH,提高了Correlation值及融合图像的清晰度,是一种有效的融合算法。The shift-variance of contottrlet transform may cause fold phenomena in fused image. Considering this disadvantage of contottrlet transform and its application in image fusion, a novel algorithm of image fusion was proposed based on Nonsubsampled Contottrlet Transform （NSCT）. First, it employs the Laplacian pyramids to achieve multi-seale decomposition and directional filter banks to achieve directional decomposition. Then it fuses the decomposition coefficients at each scale and direction. Finally, it reconstructs the fused coefficients to obtain the fused image. Due to the shift-invariance of NSCT, the new algorithm achieves a better fused image and eradicates the fold phenomena. Experimental results showed that the proposed algorithm reduced the value of MSE and △H, and increased the Correlation value as well as the sharpness of the fused image.

关 键 词：图像融合 轮廓波 平移不变性 非抽样金字塔 非抽样滤波器组 

分 类 号：V243.6[航空宇航科学与技术—飞行器设计]