Conditional Chromatic Filtering for Restoring Pansharpened Images

Pansharpening techniques aim at fusing the structural detail of the panchromatic and the color accuracy of the low-resolution spectral bands provided by Earth observation satellites to produce a high-resolution multispectral image. A large number of methods have been proposed in the literature, most of which were developed based on certain assumptions about the acquisition system. If any of these assumptions are not fulfilled, the quality of the fused products may be seriously compromised. We propose a post-processing restoration strategy to improve the quality of pansharpened images. In a first stage, we introduce a nonlocal variational model for filtering the chromatic components conditionally to the geometry of the panchromatic. In the second stage, the structural component is replaced by the locally histogram-matched panchromatic for spatial enhancement. An exhaustive performance evaluation of the proposed restoration strategy illustrates its efficiency.