中高分辨率遥感影像融合研究

以浙江省衢州市的一个区域作为研究区,在阐述遥感图像融合原理的基础上,以QuickBird、SPOT5和Landsat ETM+全色与多光谱遥感图像为数据源,应用Brovey变换、SVR变换、PCA变换、Pansharp变换和Gram-schmidt变换等融合方法,对上述3类遥感图像进行了融合试验和分析,并从光谱信息和空间信息两个方面对融合效果进行计算分析,来比较这几种融合算法用于QuickBird、SPOT5、ETM+遥感图像融合的效果。研究结果表明,在QuickBird影像融合时Gram-schmidt变换法的效果最好。Pansharp变换法在SPOT5影像融合时综合效果优于其它方法。ETM+影像融合时,Gram-schmidt变换的光谱退化最小;在空间信息增强方面:Brovey变换在波段2、3融合时效果最好,在波段1、5、7上,SVR变换法优于其它变换法,而波段4融合时则是Pansharp变换最佳。

The Study on Image Fusion for Medium and High Spatial Resolution Remote Sensing Images

With the principles of image fusion methods, this paper focused on finding a proper method for fusing the panchromatic and multi-spectral images of QuickBird, SPOT5 and Landsat ETM+. Brovey,SVR, PCA, Pansharp and Gram-schmidt were used in this paper. The results showed that within these
methods,①the optimum method for QuickBird image was Gram-schmidt transform.②Pansharp transform was the optimum for SPOT5 image.③For ETM+ image, the Gram-schmidt transform was the best method from the view of preserving spectral information of multi-spectral image; Brovey transform presented the best spatial information for band 2 and band 3, SVR transform obtained the better spatial effects than other transforms for band 1, 5, 7, and Pansharp transform was the best method for band 2.