An Improved Huber-MAP Post Filtering Method for DCT-based Compressed Video Sequences

摘要

Maximum a posteriori (MAP) post processing is widely investigated to reduce digital video or image coding artifacts. Many MAP post-filtering methods have adopted the Huber-Markov random field (HMRF) as a prior probability model. However, a drawback of the Huber-MAP post-filter is that it blurs image structures and details in video sequences when removing coding artifacts, such as blocking and ringing. This paper proposed an improved prior model to preserve edge information in images of video sequences as well as to reduce coding artifacts. Experimental results show the performance improvements of the proposed approach over the Huber-MAP post-filter. The performance of the proposed approach is also comparable to that of the H.264 post-filter, in terms of the peak signal to noise ratio (PSNR) and the visual picture quality.