可伸缩高性能视频编码的扩展技术研究

发布时间：2018-04-10 21:26

本文选题：可伸缩高性能视频编码 + 多视角联合深度信息数据结构　；参考：《中国科学技术大学》2015年硕士论文

【摘要】：近年来,随着微博和微信等新型社交媒体迅速发展,网络中视频数据量急剧增加,网络带宽和存储资源的缺口越来越大,视频压缩的重要性日益凸显。2013年1月,视频编码国际标准组织JCT-VC发布了最新一代的视频编码国际标准—高性能视频编码(HEVC)。相对于上一代视频编码国际标准H.264/AVC, HEVC编码效率提升了50%。为了满足市场中对视频各种各样的需求,在标准发布的同时,JCT-VC和JCT-3V积极推进HEVC扩展版本的研究。其中主要的扩展版本包括：支持可伸缩编码的可伸缩高性能视频编码(SHVC)、支持多视角编码的多视角高性能视频编码(MV-HEVC),支持三维视频中多视角联合深度视频(MVD)数据格式的三维高性能视频编码(3D-HEVC)。多个扩展版本能够很好地满足市场需求,但在实际应用中,可能造成用户不知如何选择和使用合适的版本。如果用统一的HEVC扩展版本能够很好地应对以上几种需求,会显著提升标准的易用性。在视频传输中,和传统的采用有状态的协议如RTSP协议相比,基于HTTP无状态的协议能够提供渐进式服务,降低了服务器和客户端的负担,提升了通信的效率,已逐渐成为市场的主流。2012年,由MPEG组织制定的基于HTTP的自适应流媒体传输技术(MPEG-DASH),能够根据网络环境和用户需求的变化动态调整多媒体资源码率,为用户提供了一个动态自适应的方法传输视频。为了支持在MPEG-DASH中的场景切换,通常需要在码流段的边界插入随机接入点。在底层编码随机接入点时,由于采用开放图片集使得场景切换点处的一些图片无法解码而产生码流中断,所以一般采用闭合的图片集的形式保证DASH场景顺利切换。本文利用SHVC编码框架的灵活性,做了两方面的研究。一方面仅仅通过高层语法的改动,使得SHVC能够较好地编码MVD数据,从而将HEVC的主要扩展版本统一用SHVC编码。另一方面提出了在MEPG-DASH中利用SHVC提升其编码性能的方法。具体来说,本文的主要工作以及创新之处在于： 1.提出了改进的SHVC编码MVD数据框架,并在此基础上提出了分量间预测,提升了深度视频和合成视频的编码性能。由于MV-HEVC和SHVC采用的都是Reference-index-based编码结构,二者可以自然统一。采用本文提出的SHVC编码MVD数据方法,可以将HEVC三个主要的扩展版本统一用SHVC编码,提高了标准的易用性。实验表明,本文所提出的分量间预测方法在深度序列编码性能和合成性能分别提升了3.6%和1.0%,很好地去除了MVD数据中纹理-分量之间的冗余。 2.提出了三种方法使得在MPEG-DASH中,利用开发图片集编码提升MPEG-DASH编码性能的同时,避免场景切换而产生码流中断。第一种方法不需要修改标准解码器,容易获取市场认同,但编码效率提升有限。第二种方法能够很好地提升编码效率,但是需要对HEVC标准解码器做简单的修改。基于此,在本文充分利用了SHVC编码灵活性基础上,提出了第三种方法—冗余自适应分辨率切换法,很好地解决了第一种方法编码效率低的问题,同时不需要修改标准解码器。因此相对于前两种方法,第三种方法更有利于市场推广和认同。实验表明,本文提出的冗余自适应分辨率切换法相对于原来的MPEG-DASH采用闭合图片集编码,平均编码性能提升了5.6%,同时解码的图片的主观质量未有明显下降。
[Abstract]:In recent years, along with micro-blog and WeChat and other new social media rapid development, a sharp increase in the amount of video data in the network, network bandwidth and storage resources gap is more and more big, the importance of video compression has become increasingly prominent in January.2013, the video encoding of JCT-VC international standards organization released a video encoding standard - the new generation of high performance video encoding (HEVC). Compared to the previous generation of video encoding of H.264/AVC international standard, HEVC encoding efficiency of 50%. in order to meet the needs of a variety of video market, in the standard JCT-VC JCT-3V released at the same time, and actively promote the HEVC extended version of the study. The extended version mainly include: support for scalable high scalable encoding the performance of video encoding (SHVC), high performance multi view video encoding support multi view encoding (MV-HEVC), support multi view 3D video and depth video (MVD) data Three dimensional high performance video encoding format (3D-HEVC). An extended version is able to meet the market demand, but in practical application, may make users do not know how to choose and use the appropriate version. If using a unified HEVC extended version can cope well with the above requirements, will significantly enhance the ease of use standard.
In video transmission, stateful protocols such as RTSP protocol and compared with traditional HTTP, a stateless protocol can provide incremental service based on reducing the server and the client's burden, improve the efficiency of communication, has gradually become the mainstream market.2012, developed by MPEG HTTP based adaptive streaming media transmission technology (MPEG-DASH), according to the dynamic changes of network environment and user needs to adjust the rate of multimedia resources, provides a method for dynamic adaptive video transmission for users. In order to support the scene switching in MPEG-DASH, usually need to insert a random access point in the stream segment boundary. At the bottom of encoding random access point. Because of the open picture set makes some pictures the scene change point cannot be decoded and stream interruption, it is generally used in the form of closed set the picture to ensure DASH scene Switch smoothly.
Using the SHVC encoding framework flexibility this paper, do the research from two aspects. On the one hand only by changing high-level syntax, so that SHVC can better encoding MVD data, which will be the main extended version of HEVC with SHVC encoding is proposed. A unified method of using SHVC in MEPG-DASH to enhance its encoding performance. On the other hand, specifically and the main work and innovations:
1. proposed SHVC encoding MVD data frame improved, and put forward the component prediction, enhance the performance of video encoding and video synthesis depth. Because MV-HEVC and SHVC are used in Reference-index-based encoding structure, two can be naturally unified. Using SHVC MVD data encoding method proposed in this paper, the HEVC can be three the main extended version use SHVC encoding, improves usability standards. Experimental results show that the proposed component prediction method in depth sequence encoding performance and synthesis performance were improved by 3.6% and 1%, very good to eliminate the redundant data between texture component MVD.
2. this paper puts forward three ways to make use of the development in MPEG-DASH, encoding MPEG-DASH encoding images to enhance performance and avoid the scene change caused interruption. Stream first method does not need to modify the standard decoder, easy to gain market recognition, but the encoding efficiency is limited. The second methods can well improve the encoding efficiency, but need to do a simple modification of the standard HEVC decoder. Based on this, in this paper makes use of SHVC encoding based on flexibility, puts forward third kinds of method of redundancy resolution adaptive switching method, a good solution to the first method of encoding the problem of low efficiency, also do not need to modify the standard decoder. Compared to the previous two methods, more third methods for market promotion and recognition. Experimental results show that the adaptive redundancy resolution switching method is discussed with respect to the original MPEG-DASH with closed Picture set coding, the average coding performance is improved by 5.6%, while the subjective quality of the decoded images is not significantly reduced.

【学位授予单位】：中国科学技术大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN919.81

【相似文献】