Effect of pre-processing on satellite image fusion

Satellite image fusion plays a key role in providing researchers with more precise and abundant information by injecting spatial details of a panchromatic image into a multispectral image. However, pre-processing of source images is required prior to the fusion process. Without well-performed pre-processing, the main fusion process cannot produce high-quality fusion results because incorrect registration of the source images can significantly degrade the fusion results either spatially or spectrally. Therefore, we analyze the effect of preprocessing on satellite image fusion in this study, using four differently pre-processed data sets. In addition, fusion framework and fusion scheme are based on the contourlet transform which outperforms the wavelet transform in terms of representing salient features of the source images. As a result of both spatial and spectral analyses, we conclude that co-registration and pre-processing of the source images can significantly improve the quality of the fusion results.