Abstract: On the basis of analysis on characters of multi-temporal image, new fusion algorithm for multi-temporal image based on change detection was proposed. First, change extent of all areas in two images was decided by combining feature extraction based on wavelet transform and ICA change detection through subspace projection, concurrence region growing algorithm was proposed in order to extract feature mask on objects. The fusion scheme was proposed based on results of change detection. Simulation results show that the algorithm proposed is very effective.

Key words: wavelet transform, independent component analysis, concurrence region growing algorithm, image fusion

摘要： 在分析多时相图像特点的基础上,提出了基于变化检测的多时相图像融合算法。该算法将小波变换的特征提取方法与ICA子空间映射的变化检测相结合,确定多时相图像各区域变化的强弱,根据本文提出的共生区域增长算法,以及变化检测的结果提出了多时相图像融合的方案,实现目标特征模板的提取。仿真结果表明,本文算法是比较有效的。

关键词: 小波变换, 独立成分分析, 共生区域增长算法, 图像融合