基于交叉累积剩余熵和NSCT的多模式遥感图像配准  

Multimodal remote sensing image registration based on cross-cumulative residual entropy and NSCT

作　　者：石永[1] 贾振红[1] 覃锡忠[1] 杨杰[2] 胡英杰 

机构地区：[1]新疆大学信息科学与工程学院,新疆乌鲁木齐830046 [2]上海交通大学图像处理和模式识别研究所,上海200240 [3]新西兰奥克兰理工大学知识工程与开发研究所

出　　处：《光电子．激光》2013年第12期2430-2434,共5页Journal of Optoelectronics·Laser

基　　金：教育部促进与美大地区科研合作与高层次人才培养(2012-1738);新疆研究生科研创新项目(XJGRI2013037)资助项目

摘　　要：图像配准是图像变化检测、融合、拼接等技术的基础,在遥感图像处理领域具有广泛的应用。利用非抽样Contourlet变换(NSCT)在图像分解上的灵活性,交叉累积剩余熵(CCRE)对遥感图像进行配准的有效性,提出一种基于CCRE和NSCT的多模式遥感图像配准算法。首先对参考图像和待配准图像分别进行NSCT分解得到低频图像,然后采用CCRE作为相似性测度,利用牛顿法获得最优仿射变换模型的参数对图像进行配准。实验结果表明,本文方法能够较快的搜索到全局最优解,配准精度高,是一种有效的配准算法。Image registration is widely used in remote sensing image processing. On one hand, non-subsamded Contourlet transform （NSCT） has the advantage of decomposing image in a flexible way;on the other hand, cross-cumulative residual entropy （CCRE） is effective in remote sensing image registration. Considering that, we propose a multimodal remote sensing image registration method which is based on cross-cumulative residual entropy and NSCT algorithm. First, the reference image and target image are decomposed with NSCT to obtain low frequency images, and then the cross-cumulative residual entropy of the obtained low frequency images is calculated. Set the cross-cumulative residual entropy as a similarri ity measurement. Secondly,Newton s method s employed to gain optimal parameters of the affine transformation model. Finally, the image registration is obtained with the optimal parameters. To validate our algorithm,we test two remote sensing images with our method. Simulation results show that the proposed method is able to find the global optimum rapidly and prevent dropping into a local minimum. In general ,it is not only a fast and effective multimodal remote sensing image registration algorithm but also the one with high registration accuracy.

关 键 词：图像配准 非抽样Contourlet变换(NSCT) 交叉累积剩余熵(CCRE) 多模式遥感图像 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]