摘要
合成孔径雷达(SAR)图像与多光谱图像成像机理和光谱特性差异较大,一般的融合方法很难取得满意的融合结果。文章提出了一种基于Nonsuosampled Contourlet transform(NSCT)和遗传算法的融合算法,首先将经过预处理后的图像进行NSCT分解,低频系数采取区域信息熵最大的准则融合;高频子带计算区域相关性,对相关性在不同阈值范围内的系数进行融合,阈值的选取采用遗传算法进行搜索;最后对融合系数进行NSCT逆变换,得到融合结果。仿真结果表明该算法显著优于基于像素点和基于区域的融合方法。
SAR image or multispectral image has big difference in imaging mechanism and spectral characteristics.Sections 1 and 2 of the full paper explain our image fusion method mentioned in the title,which we believe is new and better than previous ones.Their cpre consists of ：（1） by using the multi-scale,multi-direction ad spare decomposition capability of the NSCT,we transform the SAR and multispectral source images into NSCT domain;（2） we fuse the low frequencey coefficients by maximizing the regional entropy;then we calculate the values of the high-frequence subband regional correlation coefficients,divide them into different ranges according to the threshold values selected by genetic algorithm and fuse the correlation coefficients in different ranges,（3） we take the inverse NSCT transform,thus obtatining the fused images.Section 3 simulates our image fusion method;the simulation results,given in Fig.4 and Table 1,and their analysis show preliminarily that our image fusion method performs indeed much better than the exsiting regional-based/pixel-based Contourlet/NSCT.
出处
《西北工业大学学报》
EI
CAS
CSCD
北大核心
2012年第2期274-278,共5页
Journal of Northwestern Polytechnical University
基金
国家自然科学基金(60802084
60702063)
教育部博士点新教师基金(200806991084)
西北工业大学基础研究基金(JC20110266)资助
关键词
图像融合
NSCT
遗传算法
SAR图像
多光谱图像
analysis
algorithms
artificial intelligence
calculations
correlation methods
decomposition
design
efficiency
entropy
evaluation
frequencies
image processing
optimization
sampling
simulation
synthetic aperture radar
textures
computer software
image fusion
nonsubsampled contourlet transform（NSCT）
genetic algorithm
multispectral image
