光學衛星影像一般包括全色態影像與多光譜影像，其中全色態影像的空間解析度較高，而多光譜影像的空間解析度較低。但多光譜影像擁有較多的波段數量，可以用以描述地面物質的反射光譜，而全色態影像只有一個波段，因其頻寬較大，故可以提供較高的空間解析度。 影像融合之作用則是在於融合同一區域之多光譜影像以及全色態影像，產生出高空間解析度的多光譜影像，藉以提升多光譜影像在空間上的細節資訊。而近年來影像融合演算法被廣泛的提出，不同的演算法在融合影像上的品質有很大的差異。本研究則是提出一個新的影像融合演算法，並利用SPOT-5 和FORMOSAT-2影像進行融合實驗，再和其他演算法比較其融合影像之品質。此外本研究也針對不同的融合影像在物質分類上進行分析，比較不同演算法在分類上與原始影像分類的差異。最後針對融合影像中易出現的區塊效應 (Blocking effect) 提出解決方法。 實驗結果顯示，本研究所提出之影像融合法擁有自動化以及高效率之優點，而其融合影像不論是在視覺上以及影像品質上都有良好的表現。 Multispectral sensors on the satellites collect radiance information from the ground with several frequency bands to form a co-registered image cube. It provides useful information for spectral analysis, but it usually has low spatial resolution. On the other hand, remote sensing satellites also collect panchromatic images with wider bandwidth but higher spatial resolution. The image fusion technique can combine the panchromatic and the multispectral images into a high spatial resolution multispectral image, which can present more spatial detail of the ground. In recent years, many image fusion methods have been proposed, including fast Intensity-Hue-Situation method, Choi’s method, principle component analysis method and wavelet-based methods. However the fused images from mentioned methods have faced some problems, such as color distortion and contrast variance. In this study, we propose a new image fusion method and conduct comparison with other methods by various image quality indexes. We also apply the endmember classifier to each fused image and calculate the difference between the fused results and the original one. Finally, a deblocking method is provided to reduce the blocking effect in the fused image. In the experiments, the SPOT-5 and the FORMOSAT-2 image scenes are used .to demonstrate the effectiveness of the proposed algorithm. It provides a procedural for automatic image fusion and yields good quality and performance.
