Performance evaluation of image fusion techniques for Indian remote sensing satellite data using Z-test

Abstract

Image fusion is being used since last two to three decades in remote sensing for improving visual appearance of coarse resolution imagery using fine spatial resolution data. The resultant outputs are being used successfully in various applications such as image classification, feature extraction, digital change detection, and many more including multi-temporal and multi-scale change detection. Acceptability of a fusion method for a particular application depends upon various factors; one of them is quality of fused image. In this research paper the four different image fusion techniques namely Ehlers, IHS fusion, Brovey and FuzeGo have been evaluated using IRS-P6 (Cartosat-1) and RESOURCESAT-2 (LISS-IV) images of Bhopal city, India. Quality of fusion results is assessed by performing visual analysis between fused image and multispectral (MS) image along with statistical analysis. Visual comparison is done based on better visibility of different land cover features such as roads, buildings, water body and sharpness of edges present in image. For statistical evaluation of fusion process, six statistical parameters i.e. standard deviation (SD), correlation coefficient (CC), entropy/noise, RMSE and ERGAS have been used. In addition to these traditional statistical measures, Z-test is used for combined assessment of fusion techniques. Visual comparisons of fusion results obtained for test site have shown that FuzeGo algorithm has given comparatively better results than other algorithms. Statistical parameter CC, SD are found highest for IHS method, RMSE, and ERGAS are found highest for Brovey method and least noise is added by FuzeGo algorithm. Overall visual analysis and Z-test indicates that FuzeGo has given better results which are followed by Ehlers as compare to other methods.