节点文献

AVS解码器在PC机上的实现及优化

Implementation and Optimization of AVS Decoder on PC Platform

【作者】 弓晓霞

【导师】 张刚;

【作者基本信息】 太原理工大学 , 电路与系统, 2010, 硕士

【摘要】 我国第一个国家音视频编解码标准AVS(《信息技术先进音视频编码》系列标准简称)是数字音视频产业的基础性标准,具有自主知识产权,在技术和性能上处于国际先进水平。AVS视频编解码标准主要是为了适应数字电视广播、数字存储媒体、网络流媒体及多媒体通信等应用中大尺寸、高质量的运动图像压缩的需要而制定。Intel公司的MMXTM(多媒体增强指令集)是对Intel体系结构(IA)指令集的扩展,该技术使用了单指令多数据技术(SIMD)技术,以并行方式处理多个数据元素,可以大大提高应用程序对二维三维图形和图像的处理能力,从而提高了多媒体和通讯软件的运行速度。针对AVS视频解码算法在PC机上的实现及优化,本课题主要的研究内容及取得的成果如下:本文首先分析了AVS标准的核心技术和特点,并对多媒体指令集在视频图像领域的应用做了相关介绍。同时通过对AVS参考解码器进行一系列性能分析,结合目前主流解码器的优势,设计了符合AVS标准解码算法的C语言实现方案。然后本文重点研究了帧内解码算法的关键技术及其特点,并对其中的熵解码、帧内预测及环路滤波算法进行改进,同时编写了I帧解码器的整体代码,实现了全I帧解码器,接着对新解码器进行了测试并与参考解码模型做了详细的对比分析。经测试,在不改变图像质量的情况下,新解码效率比参考解码器提高了3倍,可实现CIF格式实时解码。最后论文探讨了MMXTM技术及其在视频压缩算法中的应用,并深入研究了MMXTM技术的编程要点。通过分析新解码器耗时的模块,确定了可用多媒体指令进行优化的部分,采用SIMD指令开发和实现了反变换、环路滤波以及帧内预测等模块。综合测试结果表明,采用多媒体增强指令优化程序,在不改变图像质量的前提下,可进一步提高解码速度。综上所述,本文研究内容几乎涵盖了在PC机上实现AVS解码算法的各个主要环节,为将来进一步完善解码器功能以及在PC机上实时实现标清和高清解码提供了有效参考。

【Abstract】 The first Chinese audio-video coding standard AVS ("Information Technology Advanced Audio and Video Coding" standards)is the basis standard in digital audio and video industry, it possesses independent intellectual property rights and has reached the international advanced level both in technique and in performance. AVS standard is established mainly for compression requirements of the moving image with large size and high-quality in the application of digital TV broadcasting, digital storage media, network communications, stream media and multimedia communication. The MMXTM technology which developed by Intel Corporation is the extension for the Intel architecture (IA) instruction set. The single instruction multiple data technology is adopted in this technology and multipe data elements is treated in parallel, which can largely improve treating performance of the application program on the two or three dimensional graphics and images, thus improving running speed of the multimedia and communication software. Aiming at the realization and optimization of the AVS Video decoding algorithm on the PC platform, the content and major research achievements in this paper are as follows:Firstly, the key technologies and features of AVS standard are analized and application of the multimedia instruction in the area of video and image are introduced in this paper. According to a serious of performance analysis on the AVS reference decoder and combining with advantages of current popularized decoder, the C language realizing scheme confirming with the AVS standard decoding algorithm is designed at the same time.Secondly, the key technology and characteristics of the intra decoding algorithm is studied emphatically, among which entropy decoding,intra prediction and loop filter algorithm is improved, integrated code of the I frame decoder is programed at the same time, full I frame decoder is realized, test on the new decoder is got, and comparison and analysis is made with reference decoder in detail. The results showed that speed of the new decoder is improved three times than that of the reference decoder, and the new decoder can realize real-time decoding.Finally, the MMXTM technology and its applications in the video compression algorithm are discussed in the paper, and programming essentials of the MMXTM technology are deeply studied. According to analyse on time consuming modules in the new decoder, parts which can be optimized based on the multimedia instruction are confirmed, and the inverse transform module, loop filter module and intra prediction module is developed and realized based on the SIMD instruction. Comprehensive testing results show that optimizing program by using the MMX can further improve decoder speed without changing image qualtiy. To sum up, the research content in this paper almost covers all main links in realizing the AVS decoding algorithm on the PC platform, which provides effective references for further performing of the decoder function and realizing of the SD and HD decoding on the PC platform.

【关键词】 AVS2D_VLDSIMD解码器实时实现
【Key words】 AVS2D_VLDSIMDDecoderRealtime Implementation
节点文献中: 

本文链接的文献网络图示:

本文的引文网络