基于Kurtosis-IHS的遥感影像融合  

Kurtosis-IHS based remote sensing image fusion

作　　者：余先川[1] 熊利平[1] 张立保[1] 徐金东[1] 

机构地区：[1]北京师范大学信息科学与技术学院,北京100875

出　　处：《地质学刊》2014年第3期380-386,共7页Journal of Geology

基　　金：国家自然科学基金项目(41272359;61071103);教育部博士点基金(20120003110032);中央高校基本科研业务费专项资金(2012LZD05)资助

摘　　要：针对传统IHS遥感影像融合的光谱扭曲和易引入噪声问题,提出了一种基于Kurtosis-IHS的遥感影像融合方法:从多光谱图像强度分量与全色图像最优逼近的角度出发,迭代求解多光谱各波段图像融合系数,减少光谱失真;根据全色图像的峭度获取其细节分量,达到抑制噪声的目的;最后使用基于Kurtosis-IHS的融合公式得到融合结果。实验选取2组实际遥感数据,结果表明,融合方法在CC、ERGAS、RASE、RMSE及SID 5个光谱性能评价指标上均优于传统的IHS、PCA、DWT及NSCT融合方法,在第二组遥感影像融合测试中,Kurtosis-IHS方法比传统IHS方法的5个指标分别提高了41.80%、57.09%、57.11%、57.11%、49.74%;在含噪环境下,融合方法的融合结果PSNR值均达到最大。所提出的Kurtosis-IHS方法优于传统的IHS融合方法,在提高空间分辨率的同时,具有较好地保持光谱信息及有效抑制噪声的优点。Due to excessive amount of information that the panchromatic image introduced in the process of fusion, traditional IHS fusion method would resulted in large spectral distortion and might easily bring extra noise. In order to overcome these two problems, the authors proposed a new IHS remote sensing image fusion method based on kurtosis. First, the authors analyzed the inner spectral distortion reason of the traditional IHS fusion, and used optimal approximation coefficients of spectral images to reduce spectral distortion. And then the authors proposed the use of kurtosis to inhibit noise. Two sets of real remote sensing image data were tested. The results showed that five spectral performance evaluation indexes of CC, ERGAS, RASE, RMSE and SID were superior to the traditional IHS methods, PCA method, DWT method and NSCT method ; and compared with traditional IHS method, the maximum values of these five indexes were increased by 41.80%, 57.09%, 57.11%, 57.11%, 49.74% respectively. The PSNR index of the study method reached its maximum in the noise experiment. Experimental results showed that the proposed Kurtosis-IHS method was superior to traditional IHS, for it improved the spatial resolution and maintained the spectral information preferably and restrained the noise effectively.

关 键 词：影像融合 全色波段影像 多光谱影像 IHS变换 峭度 

分 类 号：P715.7[天文地球—海洋科学] TP391[自动化与计算机技术—计算机应用技术]