国产高分卫星遥感影像融合方法比较与评价

选择适宜的融合方法有利于卫星遥感影像融合产品更好地服务于生产实践及科学研究。本文在总结现有影像像素级融合算法原理的基础上，选用Pansharp、Gram-Schmidt、HPF、Ehlers、Subtractive、Modified IHS、Brovey、PCA、NNDiffuse等多种常用的影像融合方法对国产高分影像的全色和多光谱数据进行了融合处理，并从定性和定量的角度对融合结果进行了详细评价，试图寻找适用于国产高分卫星遥感影像的最佳融合方法。结果表明：针对国产高分一号卫星遥感数据，超分辨率贝叶斯算法融合效果在视觉效果与影像质量定量评价指标中综合表现最佳；Gram-Schmidt、NNDiffuse、Subtractive和HPF融合结果地物边界最为清晰；Modified IHS、PCA、Brovey融合影像色彩失真较为明显；NNDiffuse在可见光波段表现较突出；Gram-Schmidt在近红外波段表现效果最佳。

Comparison of diffirent fusion methods and their performance evaluation to high spatial resolution remote sensing data of GF

Choosing an appropriate fusion method is conducive to better serving the practice and scientific research of remote sensing image. This paper tries to find the best fusion method for high resolution remote sensing images of China satellite at the basis of summarizing the existing image pixel-level fusion algorithms. Pansharp, Gram-Schmidt, HPF, Ehlers, Subtractive, Modified IHS transform, Brovey, PCA, NND and etc. methods were selected to fuse the panchromatic and multispectral data of high-resolution images of China satellites, and then 9 fusion results were evaluated qualitatively and quantitatively. The results show that Pansharping algorithm are the best both in the visual effect and the quantitative indicators of image quality evaluation. The results of Gram-Schmidt, NNDiffuse, Subtractive and HPF have the clearest boundary. Moreover, the result of Modified IHS, PCA and Brovey have obvious color distortion. The result of NND is more prominent in the visible band and Gram-Schmidt performs best in near infrared band. The information presented in this paper can provide technical reference for the application and research of GF-1 satellite data in the future.