《h.2642favc编码器的码率控制算法研究论文》由会员分享,可在线阅读,更多相关《h.2642favc编码器的码率控制算法研究论文(62页珍藏版)》请在金锄头文库上搜索。
1、华侨大学硕士论文: H.264/AVC 编码器的码率控制算法研究论文摘要ITU-T 和 ISO/IEC 两个国际标准化组织共同推出的新的视频编码标准H.264/AVC,不仅编码效率高,而且能支持从低带宽、高误码率的无线移动视频通信到高码率、低延迟的视频广播及在线流媒体等多种应用,因此日益受到业界的关注。视频压缩编码的主要目标就是在比特率受限条件下,尽量使编码失真最小。为了取得更高的压缩效率,并在接收端获得最佳的视频质量,率失真优化和码率控制技术在视频编码中占有重要地位。本文对 H.264 中的码率控制算法进行了深入的研究,针对其算法的不足提出了一些改进方法。通过对大量 QCIF 测试序列的编码
2、结果进行分析,发现不论图像类型是 I、P 帧,编码后的图像的 PSNR 值和编码量化参数 QP 具有线性关系。鉴于 I 帧占用了较多的比特数,而且其编码质量对整个GOP(图片组)的质量具有重要影响,本文利用这种线性关系来确定I 帧的量化参数 QP,以克服原 H.264 算法对于 I 帧的量化参数确定方法过于简单、不够准确的缺点,提高重建视频的质量。H.264/AVC 码率控制算法中的一个重要手段是采用线性模型来预测宏块的 MAD 值。由于它只考虑时间相关性,预测精度较低。本文综合利用视频序列中时间和空间两方面的相关性,提出了改进的MAD 预测模型,并根据前一个单元的预测误差对当前的 MAD 预
3、测值进行修正,有效地减小了预测误差。当视频序列进行场景切换时,其相邻帧间的时间相关性大大减i华侨大学硕士论文: H.264/AVC 编码器的码率控制算法研究弱,如果继续采用 MAD 值预测模型进行预测,将严重地影响码率控制的准确度及图像编码的质量。本文根据场景切换将使相邻帧图像的平均灰度差发生较大变化这个现象,快速检测是否发生场景切换,并在场景切换时对切换帧的编码模式及 GOP 结构进行调整,重新确定新 GOP 各帧的量化参数,提高了场景切换帧及同一 GOP 中后续的 P帧的质量。实验表明,本文提出的方法有效地提高了 H.264 的码率控制的性能。关键词:视频编码 H.264 码率控制 MAD
4、 场景切换ii华侨大学硕士论文: H.264/AVC 编码器的码率控制算法研究AbstractH.264 is the newest video coding standard developed by ITU-U and ISO/IEC.In addition to its high coding efficiency, H.264 also provides a solution to a broadrange of applications, such as very low bit rate conversational services andentertainment qualit
5、y broadcast, interactive video on demand services etc. Therefore,H.264 has received highly attention from academic and industrial fields. It is knownthat the main goal of video coding is to minimize the coding distortion under thegiven bit rate. In order to achieve the optimal compression efficiency
6、, and get the bestvideo quality in the receivers, rate distortion optimization and rate control technologyplays an important role in video coding. To improve the rate control performance ofH.264, this dissertation first studies the rate control algorithm of H.264 in detail, andthen proposes some imp
7、rovement methods to compensate the defect of H.264.Based on the analysis of extensive experiment results via various QCIF videotest sequences, the dissertation reveals the linear relation between PSNR andquantization parameter (QP) of a coded picture in both intra frame (I frame) and interframe (P f
8、rame). An I frame is usually allocated more bits than a P frame, and itscoding quality has more important influence to the quality of entire group of picture.However, the method used to determine the QP of I frame in original algorithm is toosimple and inaccurate. To improve the bit rate control cap
9、acity of H.264, the abovelinear relationship is introduced to determine the QP of I frame.In H.264, the original rate control algorithm employs a linear model to predictthe MAD value of a macroblock. Because it exploits only the temporal correlationamong frames, the prediction accuracy is not good e
10、nough. Hence, an improved MADpredicting model is proposed by combining the temporal and spatial correlationamong frames in this dissertation. Moreover, in this proposed algorithm, the predictedMAD value is adaptively corrected based on the prediction errors of previous unit toincrease the prediction
11、 accuracy.When the scene of a video sequence is switched, the temporal correlationiii华侨大学硕士论文: H.264/AVC 编码器的码率控制算法研究between current and previous framesis greatly diminished that the MAD predictedmodel does not work. Based on the investigation that when scene switching occurs,the average luminance d
12、ifference between two continuous frames will be much larger,in this dissertation, the average luminance difference between two continuous framesis used to fast detect whether the scene has been switched. Once scene switchingoccurs, the coding pattern of the frame and the GOP structure are adjusted a
13、daptively.Consequently, the QP of each frame in the new GOP is determined under the newcondition. Experimental results have shown that the proposed algorithm effectivelyimproves the coding quality of the scene switching frame and its following P framesin the same GOP.Experimental results have proved
14、 the proposed algorithms effectively improvethe bit rate control performance of H.264.Key Word: video coding, H.264, rate control, mean absolute difference (MAD),scene switchingiv原创性声明本人声明兹呈交的学位论文是本人在导师指导下完成的研究成果。论文写作中不包含其他人已经发表或撰写过的研究内容,如参考他人或集体的科研成果,均在论文中以明确的方式说明。本人依法享有和承担由此论文所产生的权利和责任。学位论文作者签名:日期
15、:学位论文版权使用授权声明本人同意授权华侨大学有权保留并向国家机关或机构送交学位论文和磁盘,允许学位论文被查阅和借阅。论文作者签名:签名日期:指导教师签名:签名日期:华侨大学硕士论文: H.264/AVC 编码器的码率控制算法研究第一章 绪论1.1 引言视觉是人类获取信息最为重要的途径,外部世界丰富多彩的信息大部分是通过视觉感知的。据统计,人类通过视觉获取的信息占全部获取信息的 70%1。随着计算机、数字通信、多媒体和网络技术的发展,信息也在急剧地膨胀。图像和视频作为信息最重要的载体之一,己经深入了人们的生活。人们不再满足于文字及声音,还希望看到生动的画面和场景。个人用户对多媒体业务的需求不断
16、增长,使得多媒体不再局限于文本、语音和图片,视频图像将为用户提供功能更强大、更完善的服务。近年来,视频传输的应用前景日益为人们所认识,如通过 PSTN 网络传输可视电话、无线移动网上的各种视频业务,以及互联网上的多媒体电子邮件、电子报纸、视频检索等。与语音、图片等媒体形式不同,数字视频的数据量很大,对存储器、通信信道以及处理器等都提出了相当高的要求。因此,对视频数据进行高效压缩意义重大,是降低存储成本,缓解网络带宽,突破存储空间和处理器主频限制的关键技术。码率控制(Rate Control)是视频压缩编码过程中的重要环节。它贯穿整个视频编码过程,其有效性不仅影响码流的码率稳定,还将影响到整个视频序列的图象质量。码率控制是否有效,在很大程度上取决于率失真模型(Rate-Distortion Model)的准确性及码率控制算法的适应力。同以往的 MPEG-x 和 H.26x 编码标准一样,H.264 的率失真模型及码率控制算法也没有一个统一的模式。这是因为码率控制与具体的应用需求紧密相关,不同的应用目标对码率控制的准确性、复杂性、实时性、控制灵活性、存储器资源占用量以及可控范围等方