Rate-Distortion Model And Optimization For Wireless Low-Bit-Rate Video Applications

Wireless low bit rate video service has a bright future. How to improve the quality of video service on unreliable channel and limited ability terminal, is a valuable research topic.The research on end-to-end RD (Rate-Distortion) model is very important in the field of end-to-end QoS (Quality of Service) control and RD optimization. However, the end-to-end RD characteristics depend on the application environment (i.e. the source characteristics, coding and decoding method, channel characteristics, packaging method, error correction coding, error concealment method, and so on). So, building various end-to-end RD models for various application environment is necessary. This paper studies on the GOP analysis end-to-end RD (Rate Distortion) model for the typical wireless low-bit-rate applications, and then uses this RD model to deal with two problems in practice:wireless video coding rate control and muti-users rate allocation for wireless station.The specific contributions are as follows:At First, this paper propose an end-to-end RD model for wireless low-bit-rate video coding, which is encoded into GOP structure.The model is composed of two parts:the end-to-end distortion model and the coding rate model. This model analyzes the average end-to-end distortion within one GOP and the code length of each frame in one GOP under different quantization parameters before the encoding process. Experimental results show that this model is more accurate. Compared with the RD model found on transient analysis, this model considers the overall performance of one GOP, provides the basis for bit allocation to each frame in the GOP. Compared with the RD model found on long-term average analysis, this model updates the model parameters GOP by GOP, describes the end-to-end RD dynamically.Second, as an application of the end-to-end RD model, this paper proposes a loss-aware source coding rate control algorithm.This algorithm considers the channel distortion characteristics in source coding process, improves the end-to-end video quality with the bit allocation between frames in one GOP, without changing the channel coding method or the source coding rate. Simulation results show that this algorithm preform much better than JVT-G012rate control algorithm when channel distortion exists.Compared with the earlier rate control algorithm, which considers the source distortion only, this algorithm considers both source distortion and channel distortion. In addition, this rate control algorithm is also a joint source-channel optimization method. However, compared with the traditional joint source-channel optimization strategies, this algorithm does not require cross layer design or channel coding rate control, maintains the system simplicity.Third, as another application of the end-to-end RD model, this paper proposes a multi-user rate allocation scheme for wireless station. The purpose of this scheme is to reduce the average end-to-end distortion of each user.This scheme can deal with the problem of multi-user rate allocation in polynomial time for any number of users. It adjusts the transmission rate allocation according to the dynamic channel states and video contents of each user, reduces the average end-to-end distortion of all users significantly. Simulation results have proved this point.Compared with the rate allocation scheme used in real system, which provides the same transmission rate to different users, this scheme reduces the average end-to-end distortion of all users significantly, improves the channel utilization at the same time.