Extension Of The H.264/avc Standard Based Scalable Video Coding And Key Techniques On Transmission

Scalable video coding (SVC) extension of the H.264/AVC standard has achieved significant improvement in coding efficiency when providing spatial, temporal and SNR scalability in a single bit-stream relative to the scalable profiles of prior video coding standards. SNR scalability is divided into coarse-granular scalability and medium-granular scalability (MGS). It is essential to develop transmission schemes exploiting the new features of scalable video coding extension of the H.264/AVC and the characteristics of transmission systems to enhance the performance of the whole video transmission system and improve user experience. The quality layer assignment scheme for MGS coding, unequal error protection scheme on MGS bitstream over wireless packet-erasure channel and MGS bitstream transmission framework over the emerging multi-antenna multi-user channel are investigated.Firstly, the quality layer assignment scheme for MGS coding is investigated. Bitstream with optimized rate distortion efficiency is extracted with quality layer from the encoded MGS bitstream for transmission. The current quality layer assignment scheme comes with high computational complexity. A novel quality layer assignment scheme is proposed, which comprises a reconstruction distortion estimation model and a model based quality layer computation method. The model uses only reconstruction information that is available at the encoding process to estimate the drift distortion. Its parameters can be calculated along the encoding process. Experimental results demonstrate that the proposed scheme achieves better rate distortion performance with greatly reduced computational complexity compared to the quality layer assignment scheme implemented in the joint scalable video model (JSVM).Secondly, SVC combined with unequal error protection (UEP) for delivering video robustly over wireless packet-erasure channel is investigated. Considering how to protect the bitstream with forward error correction code, a group of pictures (GOP) based UEP scheme for MGS enhacement layer packets is proposed. The novelty is that the proposed scheme takes both source selection problem and channel rate allocation problem into account. The optimization problem of UEP could be reformulated as a knapsack problem with additional constraints. A simple and effective heuristic algorithm is proposed to quickly solve the optimization problem. Experimental results demonstrate that the proposed UEP scheme provides significant error resilience compared to EEP or UEP scheme which considers channel rate allocation problem only.Finally, video transmission over the emerging multi-antenna multi-user channel is investigated. Considering the problem of SVC bitstream transmission over the MIMO broadcast channel, a framework which jointly considers the adaption of MGS bitstream and the design of zero-forcing precoder is proposed. The objective is to minimize the weighted sum of distortion perceived by all users on a time-varying broadcast channel. Given the total power, the transmitter jointly designs precoder and adjusts source rate to ensure that delay and buffer constraints are met, which is necessary for real-time video. In terms of solution approach, the framework optimization problem turns out to be a difficult mixed-integer nonlinear problem. However, it is tractable via a column generation method. To demonstrate the usefulness of the scheme, its performance is analyzed in various configurations. Simulation results show that tangible PSNR improvements can be gained compared to the typical time division multiplexing (TDM) system.