A post-processing algorithm using histogram-driven anisotropic diffusion

We propose a new post-processing algorithm for reducing coding artifacts in compressed image and video sequences. The algorithm is based on an anisotropic diffusion process using a histogram driven diffusion coefficient. The new method can adaptively adjust the diffusion degree according to the gradient distribution for different images to improve the visual quality. Experiments on JPEG compressed images and MPEG-4 compressed videos show that the proposed method can achieve consistent improvement in terms of subjective and objective quality measures.