Landsat TM and ERS data fusion: a statistical approach evaluation for four different methods

In recent years, as remote sensing imagery technology has developed, image combination from different sensors with different spatial and spectral resolutions, has become a significant technique in the field of digital remote sensing. This technique also termed data fusion or data merging is generally designed to enhance the spatial resolution of the multispectral images by combining them with high-resolution panchromatic or SAR images. Different methods have been developed to merge complementary digital data of the same area. This study try to evaluate statistically four up-to-date data merging techniques namely principal component analysis (PCA), Intensity-Hue-Saturation (IHS), Brovey and multiplicative between Landsat TM and radar SAR (ERS-1) images. Three statistical parameters have been utilized: correlation coefficient, mean and root mean square error (RMSE) in order to rank all these methods. The analysis reveals that the multiplicative method is the method distorting the least in both the original images. Paradoxically, multiplicative is the method distorting the most radar data, but the distortion is acceptable since it has a high correlation coefficient with radar data. Moreover, none of the three remaining methods have succeeded to merge Landsat TM as the multiplicative. PCA, MS and Brovey present almost the same results without significant differences.