BPG-based compression of Poisson noisy images

The article is devoted to the lossy compression of noisy images, considering the currently popular trend of green computing. In this paper, we show how this trend can be incorporated into solving an essential task of image data transferring via communication lines and what benefits it can bring to this area of interest. We consider the case of images corrupted with signal-dependent Poisson-distributed noise, which is quite common in real-life situations. Compression of noisy images has several peculiarities, among which there is the existence of optimal operation point (OOP), which is a value of compression control parameter (CCP) where a compressed image is closer to a noise-free image than the original noisy one. We show the possible presence of OOP for various visual quality metrics, both SSIM-based and PSNR-based. The recommendations for compression in OOP and its neighborhood are presented.