| In recent years, with the rapid development of the new social network like twitter and wechat, the video data increases dramatically which places greater demands on network bandwidth, recording storage, and CPU loading. Therefore, the video coding efficiency becomes more and more important. In January2013, JCT-VC released the latest generation of video coding international standards, named H.265/HEVC. Compared to the current dominant video coding standard H.264/AVC, H.265/HEVC can improve the compression efficiency more than50%. At the same time, JCT-VC and JCT-3V actively promote the extended version of HEVC to meet the market demand. There are mainly three extended verisons:SHVC for scalable extention, MV-HEVC for multiview extension,3D-HEVC for mutilview plus depth extention. Though a plurality of extented versions can satisfy the market demand well, the users may be confused which version can meet their requirement. If there has been a uniform extention of HEVC, the usability of the standard will significantly improve.In video transmission, compared to the traditional use of stateful protocols such as RTSP protocol, stateless HTTP-based protocol can provide incremental services which can reduce the burden of the server and client and improve the efficiency of communication, has gradually become the market mainstream. In2012, Dynamic Adaptive Streaming over HTTP (DASH), which can dynamically adjust the bitrate based on network environment and requirement of user, is released by MPEG. Since the open group of pictures (GOP) results into a temporary discontinuity of the video playback due to the inability to decode some pictures when switching representations, closed GOP prediction structures are normally used in DASH.This thesis proposes two works based on the scalable video coding framework of SHVC. One work proposes the improved framework for coding MVD data using SHVC without additional coding tools. The other work proposes the method which improves MPEG-DASH coding efficiency using SHVC. Specifically, the main contents and novelties are as follows:1. This thesis proposes an improved framework for coding MVD data in SHVC. Furthermore, an inter-component prediction (ICP) method is introduced to exploit the redundancy between texture and depth views. As we know, SHVC and MV-HEVC both use the so-called reference-index-based framework, they can use the same framework naturally. If the proposal is adopted, the main extention of HEVC can use SHVC framework instead. This has a positive impact on the market. The experimental results show that1.0%and3.6%bitrate reduction can be achieved by inter-component prediction compared to independent coding of texture and depth for synthesized views and depth views, respectively.2. This thesis proposes three methods for using the open GOP prediction structure in DASH representations while a full picture rate is maintained also during representation switching. The first method is easy to get market recognition as it does not need to modify the standard decoder, but the improvement of coding efficiency is limited. The second method is enabled with straightforward changes in the decoding of the HEVC standard and can own good coding efficiency. The third method, named redundant adaptive resolution change (RARC), can own good coding efficiency without any midification on the standard decoder by making full use of the adaptive resolution change feature of SHVC. As a result, compared to the other methods, the third method is more conducive to get market recognition. The proposed RARC methods outperform the use of closed GOPs by5.6%on average in terms of Bjontegaard delta bitrate (BD-rate) while there are no visual artifacts on the decoded pictures. |
PDF Full Text Request