Adaptive energy-aware free viewpoint video transmission over wireless networks
Adaptive energy-aware free viewpoint video transmission over wireless networks
复制标题
DOI：
10.1109/iccnc.2014.6785323
复制
发表时间：
2014-04
期刊：
影响因子：
--
通讯作者：
Zhi Liu;Jingyun Feng;Yusheng Ji;Yongbing Zhang
中科院分区：
文献类型：
--
作者：
Zhi Liu;Jingyun Feng;Yusheng Ji;Yongbing Zhang
随着消费级相机变得越来越便宜，现在可以由多个相机从不同的角度同时捕捉感兴趣的场景。通过传输使用两个相邻摄像头捕获的视图的纹理和深度图，用户可以使用基于深度图像的绘制(DIBR)技术(如3D扭曲)合成介于两者之间的任何视图。这为用户提供了连续的视角，是自由视点电视(FTV)等的关键技术。在这种情况下，多个视点必须通过带宽受限的无线网络传输，如何自适应地降低信源编码率以适应带宽成为一个重要问题。标准多视点视频编解码器H.264 MVC探索了视点间和视点内的相关性，以降低信源编码率。但在客户端解码任何单个视图都需要所有视图，这显然不是传输效率。更好的解决方案是使用H.264分别对相邻的两个视图进行差分编码，但与传统的视频流相比，这仍然需要双文件夹带宽。多描述编码(MDC)将一个视图编码成多个描述以实现稳健传输。这引发了另一种选择：只发送每个视图的一部分。这可以大大降低信源编码率，但在合成中间视图之前插入跳过的帧会消耗移动设备宝贵的电池寿命。是以较低的信源编码率但具有额外的能量消耗发送还是以高信源编码率发送是不重要的，并且还没有以任何正式的方式进行研究。在本文中，我们研究了在带宽约束下信源编码率的降低和能量消耗。我们将问题描述为一个优化问题，即最大化系统效用，其定义为接收到的视频质量减去加权能量消耗。仿真结果表明，在典型的网络环境下，我们提出的方案比竞争方案具有明显的优势。
As consumer-level cameras are becoming cheaper and cheaper, the scene of interest can now be captured by multiple cameras simultaneously from different viewpoints. By transmitting texture and depth maps of the views captured using two adjacent cameras, users can synthesize any view in between using depth image based rendering (DIBR) technique such as 3D-warping. This provides users with continuous view angles and is key technique for Free Viewpoint TV (FTV), etc. In this scenario, multiple views must be delivered through bandwidth-limited wireless networks, how to adaptive reduce the source coding rate to fit the bandwidth becomes an important issue. The standard multi-view video codec H.264 MVC explores the inter-view and intra-view correlation to reduce source encoding rate. But all views are required at the client side for decoding any single view, which is apparently not transmission efficient. A better solution is using H.264 to differentially encode the two adjacent views separately, but still this requires double-folder bandwidth comparing to the traditional video streaming. Multiple description coding (MDC) encodes one view into multiple descriptions for robust transmission. This inspires another option: sending only part of each view. This can greatly reduce the source encoding rate, but interpolating skipped frames before synthesizing the middle view consumes mobile devices' precious battery life. Whether to send at a lower source encoding rate but with extra energy consumption or send at a high source encoding rate is nontrivial, and has not been studied in any formal way. In this paper, we investigated both source encoding rate reduction and energy consumption with the bandwidth constraint. We formulated the problem into an optimization problem, i.e., maximizing the system utility, which is defined as the received video quality minus the weighted energy consumption. Simulation results showed that our proposed scheme provides a significant advantage over competing schemes in typical network conditions.