Evaluation and integration of markov noise reduction technique in H264/AVC

When the sequence n0i.w increases. the residue restilting from the dgerence hetween the matching macroblocks in the frame coded in predictive mode and the hest reference frame increases dramaticallv and so it i the sanie with the hit rate. One classical soliition consists in tising,/or prediction the 5 ollowed reference frames in order to find the best possihle matching macroblock. However, even with the hest match. the variation of lriminunce which reflects noise still e.rists in the residue to code. Moreover, from an implementation point of view, there would be additional processing delaw and higher memow reqiiirements. To solve this problem, we propose to embed in the H.264/A VC encoder, a technique of motion detection hased on Markov Random Fields which relies on robrrst moving pi.rel segmentation and reduces the variation of the honinance thot more ojien results,from noise rather t hmf rom an actiial motion occurring between matching macrohlocks. This algorithm needs onlv 3 reference frames to detect real niotion beween matching marroblocks and it enables decreasing the bit rate while maintaining the quality of compressed sequence. Index Terms H 2 6 4 / A V C , Markov random fields, motion estimation, noise reduction.