A pipeline to improve compressed Image Quality

This paper presents a new image restoration pipeline performing especially well on noisy and compressed images. Most images are corrupted by noise. The signal to noise ratio (SNR) level increases with the pixel intensity value, which makes the denoising process especially challenging in dark areas of the images. Moreover, these areas are more likely to be highly compressed since they have low signal variations. In this paper, we take into account compression by introducing a pre-processing step restituting the instrument noise. Then we propose a denoising and deconvolution step optimally parametrized since the instrument response (noise and Modulation Transfer Function) is known. We achieve better restoration than classical algorithms on satellite imagery. This improvement in image quality is shown on two kinds of application: pansharpening and 3D restitution.