基于方向区域的NSCT图像融合算法  

Nonsubsampled contourlet image fusion algorithm based on directional region

作　　者：周生龙[1,2,3] 张涛[1,3] 蒯多杰[3] 郑健[3] 周志勇[1,2,3] 

机构地区：[1]中国科学院长春光学精密机械与物理研究所,吉林长春130033 [2]中国科学院研究生院,北京100039 [3]中国科学院苏州生物医学工程技术研究所(筹),江苏苏州215163

出　　处：《激光与红外》2013年第2期204-207,共4页Laser & Infrared

摘　　要：提出一种新的基于方向区域的NSCT图像融合算法。算法首先对源图像进行NSCT分解,获得不同方向的高低频子带。其次对高低频系数,根据不同分解层的方向特性,按方向区域能量的规则进行融合。最后,通过反变换获得融合图像。该方法既保留了Contourlet变换方法的多尺度多方向特性,又具有移不变性。实验结果表明,提出的算法有效可行,对比常用的区域融合算法,获得了更好的融合效果。A novel nonsubsampled contoudet image fusion algorithm based on directional region is presented. Firstly, the source images are decomposed into multidirectional highpass and lowpass subbands by nonsubsampled contoudet transform. Secondly, fusion coefficients are selected according to the directional region energy. Finally, the fused image is obtained through inverse transformation. Similar to the contourlet transform method, the result of this approach is multiscalar and multidirectional image decomposition with good muhiresolution. It is also shift-invariant. Experimental results show that the proposed method outperforms the traditional transform and improves the quality of the fused im- age.

关 键 词：图像融合 无下采样轮廓波变换 方向区域能量 平移不变性 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]