遥感图像融合算法与质量评价探讨

Remote sensing image fusion methods and their quality evaluation

图像融合技术能够显著提高遥感图像的应用效益,但对于高空间分辨率卫星图像(如IKONOS、QuickBird、WorldView-2),传统融合方法往往产生光谱扭曲等现象,导致遥感图像的应用效益有所降低。笔者在总结典型图像融合算法优缺点基础上,利用五种融合算法对WorldView-2卫星图像进行融合;基于改进的综合图像融合评价指标,定量地评价各种融合算法的性能。结果表明,对于以WorldView-2为例的高分辨率遥感图像,PANSHARP算法融合的图像色彩最自然,全色波段空间细节保留完整,目标边缘锐利。

Image fusion is an integrating technique to enhance the efficiency of RS imagery. There are various high-resolution, multi-bands RS images obtained from the satellites launched after 1999 （e. g. IKONOS, QuickBird, WorldView-2 etc. ） , and conventional image fusion algorithms make more visible distortions or artifacts. Some typical fusion methods are summarized with their characters, e.g. Brovey transform, Principal Components Analysis （PCA） , Gram-Schmidt, and PANSHARP algorithm etc. In this paper, 5 algorithms are used to make fusion experiments of the WorldView-2 images; for evaluating the performance of fusion method quantitatively, a comprehensive quality index is proposed and applied to analyze the characteristics of these 5 fusion methods. The results indicat that PANSHARP algorithm is the best method to fuse WorldView-2 images.