GPGPU-assisted denoising filter generation for video coding

State-of-the-art video coding technologies such as H.265/HEVC employ in-loop denoising filters such as deblocking filter and sample adaptive offset. This paper aims to develop a new type of in-loop denoising filter using an evolutionary method. To boost the evolution, GPGPU is utilized in filtering process. Generated filter is heavily nonlinear and content-specific. Simulation results demonstrate that proposed method generates better denoising filter in 100x shorter time. The bit rate reduction of 1.492 - 2.569% was obtained against HM7.2 anchor, the reference software of H.265/HEVC.