| With the development of video compression coding technology and the establishment of related standards, digital video is more and more widely used. In recent years,with the continuous improvement of wireless network bandwidth, wireless video communications have been an indispensable part of people's life and work. However, compressed video is very sensitive to transmission errors (e.g., packet loss) and has a strict delay requirement. Furthermore, the inherent characteristics of wireless channel often result in high error rate, serious channel interference and the instability of limited transmission bandwidth. In this case, it is difficult to provide reliable video transmission quality-of-service (QoS) guarantee, which makes the QoS of wireless video transmission a challenging task.Wireless video QoS is a multidisciplinary technique involving many areas, ranging from video coding, wireless communications, and optimization theory to networking architectures, networking management and so on. Considering the guaranteed QoS is difficult to achieve for wireless video delivery, this dissertation makes an in-depth study on video encoding and decoding, resource allocation (e.g., power, bit-rate and bandwidth), packet scheduling, rate-distortion control and etc. Giving full consideration to single-channel wireless network features and video traffic characteristics, we build a reasonable objective function, and propose some novel processing methods and models to efficiently improve video delivery QoS by taking full advantage of the technologies in the related application.First of all, network packet loss rate is an important parameter for the allocation of resources. As for a given video streaming and network conditions (e.g., link status and background traffic), how to predict packet loss accurately over multi-hop network for video streaming, this work develops an artificial neural (ANN)-based packet delay bound violation model to predict packet loss probability and end-to-end distortion for video streaming over multi-hop network, and then formulates the resource allocation into a non-convex optimization frame, which aims to maximize the overall video distortion while maintaining fairness between sessions, and proposes a distributed bandwidth allocation scheme based on Lagrange dual decomposition algorithm solving the optimization problem to this end.Secondly, IEEE 802.11 physical layer supports multiple transmission rates, but does not include the rate selection strategy into protocol; therefore, how to select the best transmission rate is always a hot research issue for time-varying channel. To address this issue, this dissertation investigates the layered service mapping problem for video transmission over wireless network, and proposes an adaptive bit-rate allocation scheme with MMRP traffic model and a layered service mapping architecture that addresses cross-layer QoS issues for video delivery over wireless network. By modeling the video traffic and wireless channel as a joint Markov modulated process, and properly partitioning the states of the Markov process for real-time video transmission with QoS provisioning while achieving high channel utilization, the proposed scheme is able to provide the best tradeoff between the video application quality and the transmission capability over time-varying wireless channel.Thirdly, how to successfully encode and deliver the video data with service of a high quality over portable communication devices under the constraints of bandwidth, energy and time delay,this dissertation derives the relationship of the decoder power-rate-distortion (P-R-D) and presents a joint‘physical layer, link layer, application layer' resource allocation and performance optimization scheme. This work integrates the power consumption issue into video encoder and decoder system design. Through the adaptive adjustment of cross-layer parameters involving physical layer, link layer and application layer (e.g., video encoding and decoding), the proposed scheme allows us to find optimum energy tradeoff among video encoding, decoding and wireless transmission under resource constraints.Finally, in digital video recording and compression, the encoded bit-rate needs to be controlled so that the video storage size or transmission bandwidth constraint is satisfied. For a given bit budget, how to achieve a visually pleasing video presentation, not only does each video frame need to be encoded and decoded at the highest quality level, but also the frame-to-frame perceptual quality changes need to be smooth enough so that temporal artifacts, such as quality flicker and motion jerkiness, are minimized. To address this issue, this dissertation proposes a long-span?-domain traffic prediction model (?-LSP), by which the use of?-domain rate model is extended from the macroblock (MB)-based source rate control to frame-based traffic prediction, thus improving linearity and convergence. And then this work develops a smoothed rate control (SRC) framework based on low-pass filtering of R-D function with?-LSP and an adaptive window control mechanism for real-time video recording and streaming to address the trade-offs between playback window size and delay with buffer and channel condition constraints, controlling the online target bit-rate and video distortion effectively.Extensive experimental results demonstrate that the proposed performance optimization solutions support single-channel wireless video communications to a certain extent, so that we can find a best resource compromise between video encoding (or decoding) and wireless transmission under delayed constraint to gain significant performances. This work has achieved the purpose of combining theoretical research with practical application. |
PDF Full Text Request