Motion Vector Estimation and Adatptive Refinement for the MPEG-4 to H.264/AVC Video Transcoder
Seung-Kyun Oh, HyunWook Park · 2006
This paper proposes an efficient motion vector estimation and adaptive refinement method for the MPEG-4 to H.264/AVC video transcoder. MPEG-4 has one or four motion vectors for each 16times16 block in inter mode, while H.264/AVC has variable block sizes ranging from 4times4 to 16 times16 in inter mode in order to improve the rate-distortion (R-D) performance. The proposed method is to estimate motion vectors for various-size blocks of H.264/AVC, which exploits the decoded motion vectors of MPEG-4 for the spatial- and temporal-resolution reduction of the video transcoding. A motion vector refinement process is performed with the adaptive refinement search range determined by the directional information of the estimated motion vectors. Experimental results show that the proposed method achieves an improved R-D performance and is computationally efficient