【作者】 兰天；

【导师】 顾学迈；

【作者基本信息】 哈尔滨工业大学 ， 信息与通信工程， 2009， 博士

【摘要】 随着通信及网络技术的发展,视频通信得到了比以往更广泛的应用。在视频通信中,为了在有限的信道带宽上传输视频,需要首先对原始视频信号进行有效的压缩。码率控制是视频压缩编码的重要组成部分之一。码率控制的目标是在压缩后视频质量和编码产生的码率之间达到适合实际应用需求的平衡点。在实际系统中,码率控制需要在压缩后图像的客观质量、主观质量、实际码率、硬件限制、信道传输时延约束及特定应用需求之间做出权衡取舍。合理有效的码率控制能够在保证视频质量的同时提高信道的利用率,在有限的信道带宽下高效传输视频信息。H.264/AVC是当前视频通信中最常见的视频压缩编码标准,JVT-G012是其主流码率控制算法。JVT-G012在编码后视频质量和编码器缓冲区占用量之间取得了较好的平衡,但也有一些不足之处。针对JVT-G012的不足,相续出现了从不同角度对其进行改进的算法,如率失真模型、率失真优化、复杂度估计、缓冲区控制、信道估计、场景切换检测与处理、多序列联合码率控制及基于感兴趣区域的加权等。在本文中,以自上向下的角度,考虑到视频通信的特点,选择P帧间目标比特分配、场景切换的检测与处理及多序列联合码率控制等三个方面,进行了深入系统的研究。首先,论文针对同一GOP中各P帧的目标比特分配问题,提出了一种基于信道带宽充裕度和复杂度估计的帧层码率控制算法。采用信道带宽充裕度衡量视频序列的编码复杂度和信道带宽之间的关系。在一个视频序列编码的开始阶段判断信道带宽充裕度,并根据判断结果为码率控制中的参数集设置初始值。选择宏块级亮度直方图差异BH衡量各P帧的编码复杂度,并根据相对复杂度调整各帧的目标比特。和JVT-G012相比,针对视频序列与信道带宽的不同关系,该算法达到了不同的改进效果。当信道带宽充裕时,该算法提高了压缩后视频的客观质量与主观质量;当信道带宽不充裕时,该算法控制下的视频序列编码后产生的实际码率更接近目标码率,同时改善了编码器缓冲区占有度。其次,为降低场景切换对视频质量、信道传输及编码器缓冲区占用量的影响,论文对场景切换的检测和处理方法进行研究。分析了视频通信的特点,将满足高实时性、信道传输时延约束和编码器缓冲区尺寸约束定为设计目标。针对高实时性的需求,提出了一种基于BH的场景切换检测方法。为满足信道传输时延约束和编码器缓冲区尺寸限制,提出了一种结合前向跳帧及可变长GOP的场景切换处理方法。该算法避免了场景切换可能引发的编码器缓冲区溢出和传输延时,有效提升了新场景的视频质量。最后,针对视频通信中多序列之间的资源分配问题,论文提出了一种新的多序列联合码率控制算法。现有的多序列联合码率控制算法的控制对象是在同一编码端编码的多路电视信号,在视频通信中直接应用这些算法所付出的控制开销太高。为降低控制开销,提出了一种低复杂度的多序列联合码率控制算法。首先,该算法选择分配给各用户的信道带宽,而不是分配给各帧的目标比特,作为控制对象。其次,基于同一场景内视频序列编码复杂度变化缓慢的特点,根据已编码帧的平均视频质量调整分配给各用户的信道带宽。此外,为了应对场景切换及视频通信中的呼叫建立与呼叫终止,提出了在相应情况下的参数更新方案。和现存多序列联合码率控制算法相比,该算法同样能使不同视频序列的视频质量变得更均衡,同时该算法具有较低的计算复杂度,且能有效应对视频通信中的呼叫建立和呼叫终止,更适于应用在视频通信中。更多还原

【Abstract】 With the development of communication and network technology, video communication is more widely used in various fields. In video communication, to transfer a video sequence over limited channel bandwidth, effective video compression is essential. Rate control is one of the most important parts in video compression coding. The aim of rate control is to achieve a balance between compressed video quality and channel bandwidth to satisfy special requirements of specific application. In specific application, rate control usually need to do trade-off between objective video quality, subjective video quality, actual bit rate, hardware limitation, channel transmission delay restriction and special needs of application. An effective and reasonable rate control algorithm can not only improve compressed video quality, but also improve the channel efficiency.H.264/AVC is the most popular video compression coding standard in video communication, and JVT-G012 is the state-of-the-art rate control algorithm designed for H.264/AVC. In JVT-G012, a good balance between compressed video quality and encoder buffer fullness is achieved. However, there are still some flaws in JVT-G012. To improve the performance of JVT-G012, various improved algorithms is developed from different angles, such as Rate-Distortion model, Rate-Distortion optimization, encoding complexity estimation, encoder buffer control, channel bandwidth estimation, detection and disposal of scene change, joint rate control of multiple video sequences and Region-Of-Interest based algorithm and so on. In this dissertation, with a top-down view point, considering the features of video communication, three aspects are studied. They are target bits allocation between P frames, detection and disposal of scene change and joint rate control of multiple video sequences in video communication.Firstly, to improve target bits allocation between P frames in same GOP, a frame layer rate control algorithm based on channel bandwidth sufficiency and encoding complexity estimation is proposed. Channel bandwidth sufficiency is introduced to evaluate the relationship between channel bandwidth and the encoding complexity of video sequence. At the start of encoding a video sequence, the channel bandwidth sufficiency is judged, and the control parameters used in rate control are initialized according to the judgement. Macro-block level luminance histogram difference (BH) is applied to measure the encoding complexity of each P frame, and the target bits of each frame are adjusted according to the relative complexity. Comparing with JVT-G012, proposed algorithm achieves different improvements under different relationship between channel bandwidth and video sequence. When channel bandwidth is sufficient to video sequence, proposed algorithm improves the objective video quality and subjective video quality of compressed video sequence. When channel bandwidth is insufficient to video sequence, the actual bit rate controlled by proposed algorithm is more close to the target bit rate, and the encoder buffer fullness is improved.Secondly, to reduce the influence of scene change on compressed video quality, channel transmission and encoder buffer fullness, detection and disposal of scene change is studied. The features of video communication is analyzed, real-time processing restriction, channel transmission delay restriction and encoder buffer size restriction are confirmed as design principles. To satisfy real-time processing restriction, a BH-based scene change detection method is introduced. To satisfy channel transmission delay restriction and encoder buffer size restriction, a scene change disposal method based on forwards frame skip and variable length GOP is proposed. Proposed algorithm avoid transmission delay and encoder buffer overflow caused by scene change, and improve the video quality of new scene effectively.Finally, to solve the encoding resource allocation problem between multiple video sequences, a novel joint rate control algorithm is proposed. Existent joint rate control algorithms are designed for encoding multiple television programs at the same encoder side. In video communication, directly applying existent joint rate control algorithms will result in an unacceptable control cost. To reduce the control cost, a joint rate control algorithm with low computational cost is proposed. Firstly, the channel bandwidth allocated to each user, instead of target bits allocated to each frame, is chose as control object. Secondly, based on the slowly change of encoding complexity in the same scene, the channel bandwidth of each user is adjusted according to the average video quality of encoded frames. Thirdly, to deal with scene change and call initiation and call termination in video communication, a parameter update scheme is applied. Comparing with existent joint rate control algorithm, video sequences controlled by proposed algorithm can also achieve more uniform video quality. At the same time, proposed algorithm has an extremely low computational cost, and scene change, call initiation and call termination are effective disposed. Proposed algorithm is more suitable used in video communication.更多还原