数据同化框架下基于差分进化的遥感图像融合

针对现有融合方法的结果图像不易根据后续处理的要求进行自适应调整, 不同方法的优点不易综合的问题, 借鉴气象领域中的数据同化系统能综合其模型算子和观测算子两者优点的思想, 提出一个基于差分进化的遥感图像融合框架. 在该框架下, 将基于对比度àtrous的Contourlet变换作为模型算子, 独立分量分析和àtrous小波变换作为观测算子, 用差分进化(Differential evolution, DE)算法来优化由图像定量评价指标组成的目标函数, 从而获取更合适的图像. 二组实验从视觉效果和定量指标两方面验证了该框架的有效性.

Remote Sensing Image Fusion Based on Differential Evolution Algorithm under Data Assimilation Framework

Images obtained via existing image fusion methods could not be adjusted adaptively according to successive image processing steps and it was hard to integrate advantages of different fusion algorithms. In order to solve these problems, a remote sensing image fusion framework based on data assimilation and differential evolution (DE) algorithm was proposed in view of the advantage of data assimilation system combining the merits of its model operator and observation operator. Under this framework, contrast àtrous wavelet contourlet transform was used as the model operator, and independent component analysis and àtrous wavelet transform as the observer operator. The objective function was composed of weight sum of indices and DE was employed to obtain a proper image. Two groups of experiments have verified the feasibility of the framework in terms of both visual quality and objective evaluation criteria.