Image fusion using spatial unmixing

In remote sensing applications, image acquired from space borne satellites are of diverse spatial, spectral and temporal resolutions. Several situations in image interpretation require high spatial information and high spectral information in a single image. But the existing sensors do not have the ability to provide such information either by design or because of observational constraints. Thus the complementary information from different sensors are integrated using the technique called image fusion to get a resultant fused image which is more informative than any of the given input images. The objective of this work is to perform image fusion on a high spatial resolution LISS IV image (with low spectral resolution) and a high spectral resolution LISS III image (with low spatial resolution) to obtain a fused image with high spatial resolution and high spectral resolution. To find the endmember proportions of the mixed pixels, the land cover map is obtained by performing unsupervised soft classification on the high spatial resolution LISS IV image. The land cover class proportions thus obtained is downscaled to get the land cover class proportions at LISS III scale. The unmixing algorithm used here is spatial unmixing wherein the high spectral resolution LISS III image is processed using a sliding window approach to obtain the endmember spectra. Thus the pixels of the fused image are obtained as the linear combination of endmembers derived from the LISS III weighted by the land cover class proportions of LISS IV.