非下采样Contourlet变换医学图像融合性能研究  

Research of Medical Image Fusion Algorithms Based on Nonsubsampled Contourlet

作　　者：胡俊峰[1,2] 唐鹤云[1] 钱建生[2] 

机构地区：[1]徐州医学院医学影像学院,徐州221006 [2]中国矿业大学信息与电气工程学院,徐州221008

出　　处：《中国生物医学工程学报》2010年第4期509-516,共8页Chinese Journal of Biomedical Engineering

基　　金：江苏省高校科技成果转化项目(JHZD09-22);徐州市科技计划项目(XM09B070)

摘　　要：通过对非下采样Contourlet变换(NSCT)滤波器组合及图像融合规则性能的研究,比较了不同条件得到的融合结果,全面分析了各种融合规则对融合性能的影响。提出低通子带区域能量取大,高频子带方差加权取大和绝对值取大相结合的融合算法。相比基于传统规则的NSCT算法,所提出算法的融合质量及各项指标都有明显提高。实验结果表明,选取合适的滤波器组合和融合规则,即使分解层数较少,方向数不大,也能够获得理想的融合效果,同时还可有效减少融合算法的复杂度。Effects of the nonsubsampled contourlet transform（NSCT） were analysed and the fusion results using different filters and fusion rules were compared.An efficient fusion method was proposed fusing the low-frequency coefficients based on region energy rule and the directional high-frequency coefficients based on region variance or weighted region variance and coefficient absolute value rules.Experimental results showed the improved fusion quality compared with the traditional NSCT algorithm.Appropriate filter combination and fusion rules could bring a satisfying fusion result even with less decomposing levels and directional banks.Besides,the proposed method could reduce the fusion algorithms complexity.

关 键 词：图像融合 非下采样CONTOURLET变换 滤波器组合 融合规则 效果评价 

分 类 号：R318.04[医药卫生—生物医学工程]