隨著視訊資料編碼和網路技術的日益精進，在網路上傳輸的多媒體視訊應用日益風行。然而，MPEG-4視訊資料傳輸可能因為網路的壅塞而使得封包遺失，進而造成視訊品質下降。這篇論文主要探討在IP網路上傳輸MPEG-4即時視訊資料時，如何因應網路的流量狀況，進而控制並維持良好的視訊品質。 我們提出一套可根據IP網路壅塞現況而動態調整的MPEG-4調適性視訊品質控制機制。此機制是由編碼端的網路監控器以及解碼端的品質回授器所組成，其根據封包遺失率所對映出的網路連線狀況來動態調整編碼端之編碼輸出位元率。我們分別以畫面為基礎及物件為基礎的兩種編碼形態在各種不同網路壅塞狀況下進行調適性品質控制機制的模擬及驗證。為配合MPEG-4 VM8來源端位元率控制機制，在以畫面為編碼基礎的研究中，我們提出以調升量化參數降低位元率的方式來取代傳統的直接降低目標位元率之做法；而在以物件為編碼基礎的研究中，我們提出回授阻止錯誤蔓延機制以快速的阻擋物件的錯誤蔓延，並利用物件可調性觀念進行視訊品質控制，保護重要物件之傳輸優先權。實驗結果顯示，不管是以畫面或物件為基礎的情況下，在網路嚴重壅塞時（封包遺失率為20%），加入所提之品質控制機制的平均PSNR比沒有品質控制提升1.27到2.16 dB。 The MPEG-4 VM8 source rate control use both the frame-based and multiple video object (MVO)-based encoding methods to generate video sources toward a target rate. To incorporate the network conditions, we propose an adaptive MPEG-4 quality control mechanism to overcome various congestion situations of IP network for maintaining the quality of MPEG-4 video source. The mechanism composes of network monitoring unit (NMU) at the encoder end and quality feedback unit (QFU) at the decoder end. Based on the network condition represented by the packet loss rate, the mechanism adjusts the encoding output bitrate dynamically. We verify and simulate the proposed adaptive quality mechanism under various congestion conditions. In frame-based adaptive quality control the frame output bitrate is adjusted by changing the value of quantization parameter instead of decreasing the overall target bitrate directly. In MVO-based adaptive quality control we use feedback to stop error propagation of MVO shape coding rapidly. Moreover, we utilize the object scalability property to protect the significant objects by high priority transmission. Simulation results show that either frame-based or MVO-based adaptive quality control mechanism improves the average PSNR more remarkably than no quality control under the heavy congestion network condition with a packet loss rate 20%. The improvement of average PSNR is in the range from 1.27 to 2.16 dB.
