Compressive Sensing based Satellite Image Fusion

Abstract

Image fusion is the process of integrating the relevant information from two or more images into a single image. In traditional methods of pan sharpening, the fused images suffered from spectral distortion. In recent literature, filter based methods were proposed and they are often used as compared to other fusion methods because of their ability to improve spatial and spectral quality. In filter based techniques, spatial information from panchromatic image is injected into the multispectral image. A compressive sensing based filter approach is proposed in this research. Compressive sensing aims to construct the fewer samples from many signals by using only a few non-zero coefficients in a suitable basis. These few samples can efficiently represent the input signals. In proposed technique, regression coefficients are calculated from compressive signals of panchromatic and multispectral images. An optimized filter is constructed from regression coefficients and input images. Filter extracts the significant and non-redundant information from input images. Low resolution panchromatic image is then constructed from optimized filter and input panchromatic image. Finally, low resolution panchromatic image is injected into the multispectral image. Input images of different resolutions, sizes, backgrounds and locations were used in our experiments. Final fused images were evaluated using two spectral quality measurement techniques and four fusion quality assessment techniques. Visual and quantitative results show that the proposed algorithm produces better results in terms of spectral and fusion quality as compared to the previously presented methodologies.