Efficient MMSE pansharpening based on non-local optimization

The paper presents a pansharpening algorithm that finds an optimal linear solution, in the MMSE sense, following a generalized component-substitution approach. It is characterized by nonlocal parameter optimization obtained through K-means clustering. The proposed method, namely C-BDSD, solves the problem of context-adaptive schemes that tune the spatial injection parameters on local statistics: instabilities and blockiness artifacts are avoided and the estimation phase is improved. The C-BDSD algorithm is accurate and fast, and can be also applied to spatially enhance large-size multi-spectral images. Very high quality scores and excellent visual quality of the fused images demonstrate the validity of the method.