壓縮視訊在封包網路上的傳輸

Abstract

一般而言，可變速率編碼器在封包網路環境較固定速率編碼器易達到固定 的影像品質和較高的網路效益。然而，封包網路也帶來固定速頻道所沒有 的新問題，例如封包遺失和延遲。經由影像編碼器控制參數的選用來追求 較高網路使用效益是這篇論文的目標。我們分三個部份討論。在第一部份 ，我們考察在時域作切割的雙層編碼器，並評估不同調節輸出方式的功效 ，將統計結果于以收集分析。第二部份是從整個系統的觀點探討最佳的操 作點。整個系統包括視訊編碼器和網路。終極的目標是減少接收端重建影 像的總失真量，這失真包含了信號源編碼誤差和頻道誤差。我們模擬的結 果顯示此最佳點的存在。最後介紹一個輸出位元調節系統，它可以減低視 訊編碼器所產生位元數的振盪。如果變動位元率沒有任何限制而直接進入 網路，將易造成網路壅塞而有封包遺失。影像的品質是隨封包遺失量的增 加而降低。但如果變動位元率進入網路之前，經過我們提出的架構調整， 封包遺失的數目將會減少很多，而達到較高的網路效益。 Variable bit rate coders can in general achieve constant picture and higher network efficiency than the constant bit rate coders for packet network communication. However, packet networks also introduce new problems that are not encountered in the constant bit rate channels, such as packet loss and delay. The goal of this thesis is to achieve higher network utilization. It consists of three parts. In the first part proposes a two-layer coding scheme is introduced. We investigate a temporal partitioned two-layer codec structure and different modes for transmission. The second part explores the optimal operation point in the overall system including both video codecs and transmission network. Our ultimate goal is reducing the overall distortion on the reconstructed images at receiver and this distortion contains the source coding error and the channel loss error. The last part introduces an output regulation that smoothes the fluctuation of output variable bit rates generated by video codec. If variable bit rate is regulated by this scheme before entering to network, the number of lost cells would reduce significantly to achieve higher network efficiency.