Research On Rate-Distortion Optimization Based Rate Allocation In Scalable Video Coding

With the rapid development of network and mobile communication in recent years, the target of video coding changes from traditional requirement for storage to network transmission. Scalable video coding can generate bit-stream with spatial, temporal and SNR scalability, which makes video signal transmitted robustly over the isomerous network with time-varying bandwidth. Due to the fluctuating of bandwidth or different terminal units, fast and efficient rate allocation technology must be taken in order to get an optimal bit-stream under a certain condition. In terms of these requirements, we carry on research of rate allocation in scalable video coding.Embedded Block Coding with Optimized Truncation (EBCOT) is a core algorithm in JPEG2000. It can compress image efficiently and produce bit-stream with strong scalability. JVT is currently standardizing scalable video coding (JVT SVC) which includes many advanced scalable technologies as an extension of excellent video coding standard H.264/AVC. Therefore, we first investigate EBCOT in JPEG2000 and JVT SVC, especially their rate allocation technology, which are valuable references to the design of new rate allocation methods in this thesis.Then, we notice that rate-distortion information is calculated for every FGS NAL unit after examining existing method of rate allocation in JVT SVC, which is not precise enough to describe rate-distortion characteristic. To eliminate defect of existing method, an optimal rate allocation method is proposed. First, an approximate rate-distortion (R-D) curve of each FGS NAL unit is established by dividing FGS NAL unit into several fragments. Then, the R-D points lying on the convex hull of curve are chosen as R-D optimization truncation points, according to which rate allocation is optimized. Experimental results indicate that when extracting bit-stream with the same requirement, compared to the existing method of R-D optimized rate allocation in JVT SVC, quality of decoded sequence is further improved.Finally, we found that the impact distortion of one FGS NAL unit on the global reconstruction quality is calculated by repeating decoding sequence as existing method of rate allocation in JVT SVC, which leads to high complexity. Hence, we present a fast method of rate allocation by modelling distortion estimation. Base on the hypothesis that residual coefficients confirm to Laplacian distribution, we establish a model by calculating quantization error and analyzing the prediction relationship in hierarchical B coding structure. Meanwhile, the parameter in the model is updated according to the distribution of residual coefficients every frame. Experimental results show that the proposed method can obtain a reduction in decoding time of nearly 50%, and save the runtime of rate allocation by 45.3%, while the PSNR loss of decoded sequence is only 0.04 dB on average. Thus, it can be concluded that the proposed fast method has achieved fast and effective rate allocation.