基于数据同化和遗传退火的多聚焦图像融合  

Multi-focus Image Fusion Based on Data Assimilation and Genetic Simulated Annealing Algorithm

作　　者：陈荣元[1,2] 张飞艳[1] 李爽[1] 秦前清[1] 

机构地区：[1]武汉大学测绘遥感信息工程国家重点实验室,武汉430079 [2]湖南商学院信息学院,长沙410205

出　　处：《计算机辅助设计与图形学学报》2009年第8期1134-1137,1143,共5页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家“九七三”重点基础研究发展计划项目(2006CB701303);国家“八六三”高技术研究发展计划(2006AA12Z132);湖南省教育厅资助科研项目(08C485);长沙市软科学项目(K0802190-41)

摘　　要：针对现有的融合方法不能根据融合图像的后续使用目的对融合规则进行调整的问题,提出一个基于数据同化和遗传退火算法的多聚焦图像融合框架.该框架将小波变换作为模型算子,把主成分分析法作为观测算子,根据后续处理对图像各个属性指标值的依赖程度确定各个属性指标的权重;再用各个评价指标的加权和来构造目标函数;利用遗传退火算法优化目标函数,以获取更合适的图像.最后通过一组实验证明了该框架的有效性.The existing image fusion methods usually set down the fusion rules before fusion processes. However, the rules which determine the attributes of fusion results could not be adjusted according to different future applications. In order to overcome this limitation, a framework for multi- focus image fusion is proposed based on data assimilation and genetic annealing algorithm. Under this framework, the wavelet transition method is regarded as the model operator, while the principal component analysis method is considered as the observation operator. The weights of different attributes are determined according to their different importance to the post-processing. The weighted sum of the evaluation indices is used to construct the object function. Finally, the object function is optimized by using genetic annealing to obtain the proper image. The experiments validate the effectiveness of the framework.

关 键 词：数据同化 遗传算法 退火算法 图像融合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]