【作者】 顾锋；

【导师】 孙军；

【作者基本信息】 上海交通大学 ， 信号与信息处理， 2012， 硕士

【摘要】 视频网络通信对于带宽,延时与传输可靠性有很强的要求。相对于传统二维视频,立体视频包含更多视角的信息,必然会造成数据量成倍地增长,因此立体视频必然对网络通信带宽提出了更高的要求。因为IP网络尽力而为的设计机制,导致视频网络通信过程中不可避免的会发生数据包丢失。立体视频具有多个视角信息,不同视角之间具有相关性。一个视角数据发生损伤会影响别的视角进而影响整个立体视频的视觉质量。因此降低立体视频网络通信所需的带宽与网络损伤程度是保证立体视频视觉质量的关键问题。为了更契合立体视频网络通信的要求,本文针对立体视频对带宽与网络损伤的要求,着重研究了立体视频网络传输过程中的码率平滑以及网络损伤与压缩比例对立体视频视觉质量影响的关键问题。立体视频视角可分级传输过程中各视角信道码率会因周期性突发传送而出现明显的抖动现象,码率抖动幅度达到100%以上,增加了立体视频网络传输所需的带宽,严重影响网络传输质量。本文针对立体视频视角可分级传输方法各视角码率抖动的特点,提出了自适应双缓冲码率平滑策略。通过在发送端对每个视角进行独立缓冲码率平滑,降低了立体视频码率抖动的幅度,将码率抖动控制到10%以内。立体视频因为具有多个视角信息,数据量会成倍的增长,而IP网络并不能保证稳定的带宽供应。为了降低立体视频网络通信的难度,就要尽可能的压缩视频信息,使用较少的数据量来表示较高的立体视频视觉质量。高度压缩的立体视频数据包具有强烈的相关性。数据包的丢失会影响整个立体视频的视觉质量。为了得到在不同的压缩比例的条件下,经过不同程度的网络损伤后立体视频的视觉质量,本文使用MPEG2编码标准、采用4:2:0的采样格式对分辨率为1920×1080的side-by-side格式的立体视频进行不同比例的压缩编码,然后将视频数据包经过不同丢包率的网络进行损伤仿真,最后统计出损伤后立体视频的主观感觉质量。实验得出1920×1080的side-by-side格式的立体视频在码率达到8Mbps-9Mbps、丢包率在0.0002-0.0005之间时,立体视频的视觉质量才能够满足用户的主观要求。实验最后对比得出在相同参数条件下,side-by-side格式的立体视频的主观视觉质量都要次于二维视频,立体视频对网络通信质量需要更高的要求。更多还原

【Abstract】 The transmission of video over IP network has a strong demand on bandwidth, delay and reliability. More views mean more data. Comparing with traditional single view videos, Multi-View videos would occupy more transmission bandwidth. Because of the scheme of do one’s best, packet loss is inevitable during the video transmission over IP network. When packet loss happens to one view of multi-view video, it will have a bad effect to other views and to the whole video. Thus, reduce the transmission bandwidth and degree of damage is the key point to ensure the quality of multi-view video.In order to satisfy the demand of transmission of multi-view video, the essay focus on the bandwidth and the damage of transmission of multi-view video, studies the rate smoothing scheme and the effect of damage to multi-view video.In the view classification transmission scheme, data from different views are transferred in a circle order, which would cause a drastic jitter on the transmission rate of each view, the phenomenon will require more transmission bandwidth which would seriously affect transmission quality. According to the characteristic view classification transmission scheme, the author introduce an adaptive double-buffered rate smoothing method and managed to restrict this jitter to less than ten percent by applying the rate method to each view.Because of multi angles, the data amount of multi-view video will several times larger than traditional single view video. But IP network can’t supply steady bandwidth. In order to reduce the difficulty of transmission of multi-view video, we need to compress video information as far as possible to show multi- view video with less amount of data. The data packet of highly compressed multi-view video has a strong relativity with each other. Each packet loss will affect the whole visual quality of multi-view video. This essay encode the video in resolution 1920×1080 with MPEG2 standard in 4:2:0 sample format, then send the video through network with different qualities. It turns out that while the bit rate of side-by-side multi-view video achieves 8Mbps-9Mbps and at least the packet loss rate is between 0.0002-0.0005, the video quality will satisfy user’s subjective feelings.更多还原