Research On Rate Control In Video Transmission

In multimedia communications, compressed video streams are transferred via networks with time-varing bandwidth. In order to make full use of network resource, and guarantee the best visual quality of users, Multimedia communication systems should adopt some flow control mechanism. Supported by the national high technology research and development program of China (863 Program)—Joint Source-Channel Coding (No.2002AA119010) and Digital Audio-Visual Coding technology workgroup(AVS workgroup), this thesis researches the rate control schemes of emerging audio-visual coding standards, especially the newly standard H.264/AVC Part10.The main work of this thesis is to put forward better rate control schemes than those in H.264/AVC Part10. Although H.264 has brought forward a series of efficient video coding methods, it does not do much research work on rate control. It uses quantization parameter both in rate control and R-D optimization, resulting in the chicken-egg dilemma. Thus it is difficult to realize an efficient rate control scheme in H.264. This thesis researches the two efficient rate control schemes ——JVT-G012 and JVT-F086 of H.264, and put forth two different rate prediction models and corresponding rate control schemes: a frame layer scheme and a macro-block layer scheme. Simulation results show that, the two schemes exceed JVT-G012 and JVT-F086 schemes in control precision, rate variety, visual quality resilience and robustness. What's more, this thesis introduces the international multimedia conference standard H.323 and multimedia conference system OpenH323. As a system of real-time distance conference, OpenH323 is suitable for large scale distance learning and conference in school,government,cultivate organization and big corporations. With the realization of H.264, the performance of OpenH323 will be enhanced greatly. In addition, this thesis introduces the research and improvement on OpenH323 by rate control: after improving OpenPhone, the visual quality enhances greatly and the congestion of networks is probably avoided; with the exploitation of OpenMCU, the maximum visual number is augmented; by exploiting OpenGK, the system information increases greatly. Simulation results show that the improvement on OpenH323 is very efficient.