Comparative statistical-based and color-related pan sharpening algorithms for ASTER and RADARSAT SAR satellite data

In computer vision, multi-sensor image fusion is the process of combining relevant information from two or more images of a scene into a single composite image. The resulting image will be more informative than any of input images. In this study, the efficiency of different pixel-based Pan sharpening techniques for merging RADARSAT-1 SAR and ASTER-L1B data is investigated and compared. In doing so, two statistical-based techniques including the Gram-Schmidt and Principal Component transforms, and two color-related techniques including the Brovey and HSV transforms are applied to merge the satellite images. One of the major problems associated with data fusion techniques is how to assess the quality of the fused images. In this regard, several indicators such as the Relative Mean Difference (RMD), Relative Variation Difference (RVD), Root Mean Square Error (RMSE) and Spectral Quality Indices (SQI) are used to evaluate the performance of the fused images. Then the fusion techniques are ranked according to the conclusion of each indicator. The achieved results from the relative mean difference analysis indicated advantage of the PC and GS than the Brovey and HSV transform techniques. The results based on relative variation difference and root mean square error indicated superiority of the PC transform while the results of spectral quality indices showed advantage of the GS transform technique. The output of HSV transform indicated the worst result and disadvantage of this technique in all indicators. In conclusion, it can be said that the PC is the best, the GS is better, the Brovey is bad and the HSV is the worst technique for multi-sensor data fusion. Finally, all indicators indicated advantage of the statistical-based fusion techniques than the color-based to fuse the ASTER-L1B and RADARSAT-1 SAR data.