Fast mode decision algorithm for spatial resolutions down-scaling transcoding to H.264  

作　　者：BU Jia-jun MO Lin-jian CHEN Chun YANG Zhi 

机构地区：[1]School of Computer Science, Zhejiang University, Hangzhou 310027, China School of Computer Science, Zhejiang University, Hangzhou 310027, China School of Computer Science, Zhejiang University, Hangzhou 310027, China School of Computer Science, Zhejiang University, Hangzhou 310027, China

出　　处：《Journal of Zhejiang University-Science A(Applied Physics & Engineering)》2006年第z1期70-75,共6页浙江大学学报（英文版）A辑（应用物理与工程）

基　　金：Project supported by the National Natural Science Foundation of China (No. 60573176)the Key Technologies R & D Program of Zhejiang Province (Nos. 2005C23047 and 2004C11052), China

摘　　要：A fast mode decision algorithm is proposed in this paper to accelerate the process of transcoding videos into H.264with arbitrary rate spatial resolution down-scaling. The proposed algorithm consists of three steps. First, an early-stop technique is introduced to determine the 16× 16-mode blocks, which take up about 70% of all the macroblocks; then, a bottom-up merging process is performed to determine the mode of rest non-early-stopped blocks; and then, we adopt half-pixel motion estimation to further refine the acquired predictive motion vectors. In order to obtain the predictive motion vectors for early-stop and merging processes, we propose a motion vector composition scheme, which can reuse the information in the input pre-encoded videos to handle the spatial resolution down-scaling. Experimental results showed that our algorithm is about four times faster than the Cascaded-Decoder-Encoder method and has negligible PSNR drop and little bit rate increase.A fast mode decision algorithm is proposed in this paper to accelerate the process of transcoding videos into H.264 with arbitrary rate spatial resolution down-scaling. The proposed algorithm consists of three steps. First, an early-stop technique is introduced to determine the 16×16-mode blocks, which take up about 70% of all the macroblocks; then, a bottom-up merging process is performed to determine the mode of rest non-early-stopped blocks; and then, we adopt half-pixel motion estimation to further refine the acquired predictive motion vectors. In order to obtain the predictive motion vectors for early-stop and merging processes, we propose a motion vector composition scheme, which can reuse the information in the input pre-encoded videos to handle the spatial resolution down-scaling. Experimental results showed that our algorithm is about four times faster than the Cascaded-Decoder-Encoder method and has negligible PSNR drop and little bit rate increase.

关 键 词：Transcoding  H.264  Mode decision  Motion composition 

分 类 号：TN919.8[电子电信—通信与信息系统]