Sparse representation based pansharpening with details injection model

Sparse representation based pansharpening attracts a lot of interests, which formulates the pansharpening as a compressive sensing reconstruction problem. In this paper, a new pansharpening method based on sparse representation with details injection model is proposed. Differently from the existing sparse representation based methods, the proposed method adopts the ARSIS concept instead of the compressive sensing reconstruction, and the pansharpened image is created by the details injection. Therefore, the prior assumption is not required to define the reconstruction model. The proposed method applies the sparse representation based image decomposition to extract the spatial details from the panchromatic image. Meanwhile, only the panchromatic image in the a trous wavelet domain is used to learn the dictionary that makes the proposed method more practical. The quantitative results and visual evaluation on the QuickBird and WorldView-2 data show that the proposed method is comparable or even superior to the existing sparse representation based methods and some conventional methods. (C) 2014 Elsevier B.V. All rights reserved.