Application of Adaptive Injection Gain in Sparse-Based Multispectral Image Fusion

In this paper, an adaptive procedure for determining the injection gains in a model-based pansharpening method is proposed. The fusion process is done in a patch-wise manner by adding the detail patches to the low resolution multispectral (LMS) patches. The detail patches are obtained from a sparse linear combination of the atoms (columns) of the detail dictionary which is constructed from the high-frequency information of the panchromatic (PAN) image. Therefore, the correlation coefficient between the PAN image and the ideal high resolution multispectral (HMS) image is considered as the injection gain of the details patches. To address the problem of unavailability of the ideal HMS image, an iterative algorithm is proposed which adaptively determines the injection gains. Also, the weights of constructing the intensity component, which is used to find the sparse coefficients of the detail patches, are optimally calculated using the least square error method. The simulation results of the Pleiades and GeoEye-l data demonstrate the superiority of the proposed method in comparison with the state-of-the-art methods in both visual and quantitative aspects.