The Research Of Macroblock Mode Decision For Spatial Resolution Transcoder Based On H.264

With the development of the computer science and multimedia communication,several video coding standards have been established for various video applications. As the number of networks and the coexistence of the video standards, interoperability between different systems and different networks is becoming more and more important. Video transcoding is an efficient way to solve this Problem. Video transcoding contains two kinds of transcoding problems: homogeneous transcoding and heterogeneous transcoding. The homogeneous transcoding, which can be claasified as spatial resolution transcoding, temporal resolution transcoding and bitrate transcoding.In H.264, Variable macroblock mode is adopted which leading to high compression capability and high computation complexity as well. In spatial resolution transcoder based on H.264,the pre-encoded macroblock modes are not suitable anymore. A fast macroblock mode decision method is proposed in this paper in arbitrary downsizing video transcoder based on h.264. In spatial resolution transcoder with arbitrary factor, the responding area in pre-encoded frames for macroblocks to be re-encoded is not aligned with the boundary of macroblock. The concepts of mapping block and covering block are proposed in this paper. To reduce the computation complexity, method based on blocks or based on pixels is adopted according to different downsizing factor. In inter-prediction mode decision part, SKIP mode is analyzed specially. The percentage of SKIP mode in pre-encoded frame is considered in the proposed scheme to decide SKIP mode in frame to be re-encoded. We compared the PSNR performance and computation complexity of proposed method with the full search algorithm by extensive experiments.For the intra-frame mode decision part,the proposal achieves about 3.5 times improvement in computation complexity comparing to full search algorithm while the maximum PSNR is degraded by about 0.1dB; For the inter-frame mode decision part, the proposal achieves about 30 times improvement in computation complexity while the maximum PSNR is degraded by about 1.0dB.