Fast mode decision and rate control for H.264/AVC and SVC extension

視頻編碼的快速算法和碼率控制

Student thesis: Master's Thesis

View graph of relations

Author(s)

  • Sudeng HU

Related Research Unit(s)

Detail(s)

Awarding Institution
Supervisors/Advisors
Award date4 Oct 2010

Abstract

In this thesis, a Fast Inter-Mode Decision algorithm is proposed for H.264/AVC and Rate Control(RC) algorithms are proposed for temporal and spatial layer Scalable Video Coding (SVC) respectively. Firstly, a new fast mode decision (FMD) algorithm is proposed for the state-of-theart video coding standard H.264/AVC. Based on Rate-Distortion (RD) cost characteristics, all inter modes are classified into two groups, one is Skip mode (including both Skip and Direct modes) and all the other inter modes are called non-Skip modes. In order to select the best mode for coding a Macroblock (MB), minimum RD costs of these two mode groups are predicted respectively. Then for Skip mode, an early Skip mode detection scheme is proposed; for non-Skip modes, a three-stage scheme is developed to speed up the mode decision process. Experimental results demonstrate that the proposed algorithm has good robustness in coding efficiency with different Quantization parameters (Qp) and various video sequences and is able to achieve about 54% time saving on average while with negligible degradation in Peak-Signal-to-Noise-Ratio (PSNR) and acceptable increase in bit rate. Secondly, for temporal scalable video coding, a novel frame-level RC algorithm is presented in this thesis. By introducing a linear quality dependency model, the quality dependency relation between a coding frame and its references is investigated for the hierarchical B-picture prediction structure. Linear Rate-Quantization (R-Q) and Distortion-Quantization (D-Q) models are introduced based on different characteristics of temporal layers. According to the proposed quality dependency model and R-Q and D-Q models for each temporal layer, an adaptive weighting factor is derived to allocate bits efficiently among temporal layers. Experimental results on not only traditional QCIF/CIF but also Standard Definition (SD) and High Definition (HD) sequences demonstrate that the proposed algorithm achieves excellent coding efficiency as compared to other benchmark RC schemes. Thirdly, for spatial layer of Scalable Video Coding, a novel rate control algorithm is presented in this thesis. A new best initial Qp model is proposed based on the power R-Q model. By applying the proposed sequence complexity measurement, the proposed model can provide proper initial Qp before encoding. Then the relationship between the best initial Qps of different layers is investigated and determination of the best initial for multiple Qps layer is introduced. Meanwhile by introducing a two stage RC scheme, a novel frame complexity estimation method is proposed. The dependency of the parameters in the RQ model is investigated to improve the model accuracy. The experimental results demonstrate that the proposed RC scheme and best initial Qp perform excellent coding efficiency and accurate bit achievement.

    Research areas

  • Video compression, Coding theory, Digital video