一种多极化SAR图像融合方法研究  

Fusion algorithm for multi-polarization SAR images based on Nonsubsampled Contourlet Transform and PCNN

作　　者：杨智翔[1] 何秀凤[2] 危来龙[1] 杨宁[1] 王林[1] 

机构地区：[1]江西省水利规划设计院,南昌330029 [2]河海大学地球科学与工程学院,南京210098

出　　处：《测绘科学》2014年第2期14-18,共5页Science of Surveying and Mapping

基　　金：国家自然科学基金(41274017);江苏省科技支撑计划(BE2010316);日本宇航局ALOS数据研究项目(PI 534)

摘　　要：本文针对多极化SAR图像的融合问题,提出了一种基于非下采样Contourlet变换(NSCT)与脉冲耦合神经网络(PCNN)的图像融合方法。此方法用NSCT对已配准的多极化SAR图像进行分解,得到低频子带系数和各带通子带系数;采用简化的PCNN模型分别对低频子带和高频子带系数进行智能决策,并进行NSCT逆变换得到融合图像。经实验表明该方法能够最大程度地保留原始极化SAR图像的信息,融合效果好于基于单个像素和局部特征的融合方法。Considering the limitations of the traditional fusion rules based on pixel or local characteristics, Pulse Coupled Neural Network （PCNN） with the global coupled property was introduced to image fusion. Combining with the excellent characteristics including multi-scale, multi-direction and shift-invariant in the Nonsubsampled Contourlet Transform（NSCT）, a new fusion scheme based on NSCT and PCNN was proposed to fuse multi-polarization Synthetic Aperture Radar （SAR） images in the paper. The simplified PCNN model was used to make decision on the sub-band coefficients selection in NSCT domain. Final ly, the method was examined by using ALOS dual-polarization SAR images and compared with some regu lar fusion algorithms based on multi-scale decomposition. Experimental results indicated that the proposed method would be more effective to fuse the multi-polarization SAR images than the pixel-based algorithm and the windows-based algorithm.

关 键 词：非下采样CONTOURLET变换 脉冲耦合神经网络 多极化SAR图像 图像融合 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]