Panchromatic and multispectral images fusion using sparse representation

In this paper, we propose a new pansharpening method based on sparse representation theory to fuse panchromatic and multispectral images. In the proposed method, the high-resolution multispectral image is reconstructed by adding some details to the multispectral image. The details are achieved directly by a proper dictionary which is constructed using a high pass version of the panchromatic image, so-called ‘detail dictionary’, and proper sparse coefficients. The required atoms for generating the details are chosen by two objective functions. One of these functions chooses atoms having high spatial information and the other one selects atoms with high spectral information. Then, the details are made from a linear combination of these atoms. We use both sets of the atoms to increase the spatial details and decrease the spectral distortion. In order to investigate the efficiency of the proposed method, two datasets from Pleiades and WorldView-2 satellites are used. Based on the experimental results, it is found that the proposed method performs better than the state-of-the-art methods in maintaining of spectral information as well as increasing spatial details objectively and visually.