應用傳輸錯誤更正具精細伸縮特性之視訊可靠多點傳輸

Abstract

即時影像傳輸在眾多的網際網路多媒體應用中，是一個非常重要的組成要件。視訊串流由於其即時傳輸的特性，對網路頻寬、封包延遲以及封包遺失必須有特定的要求。然而，現今的網際網路對於視訊串流並不提供服務品質的保證，因此，為了達到上述的需求，我們必須把壅塞控制以及錯誤控制列入考量。 壅塞控制的目的是為了預防封包遺失的發生，而錯誤控制的目標則是恢復封包遺失的傷害。為了達成壅塞控制，我們採用具精細伸縮特性的新影像壓縮技術，其可以支援視訊串流伺服器動態調整視訊資料傳輸率以因應網路頻寬變化，並且對於封包遺失也具有恢復能力。此外，由於多點傳輸是一種兼顧效率以及彈性的傳輸方式，我們決定使用這個技術來傳送視訊資料。 然而，錯誤控制在多點傳輸的模式中仍然需要考慮。在這篇論文中，我們展現一個可靠的多點傳輸視訊串流系統－應用傳輸錯誤更正具精細伸縮特性之視訊可靠多點傳輸。這個系統的特色包含(1)一個可以支援封包遺失回復的應用層通訊協定(2)一個能快速存取視訊資料以及幫助封包遺失回復的緩衝區架構(3)為了達成縮減封包遺失回復時間以及降低封包遺失回復頻寬的整合重送與傳輸錯誤更正可靠機制。 最後，我們展示系統的效能來證明它的確在及時影像傳輸中提供了高度的可靠性。

Delivery of real-time video over the Internet is an important component of many Internet multimedia applications. Due to its real-time nature, video streaming typically has special requirements on bandwidth, delay and loss. However, the current best-effort Internet does not offer any quality of service (QoS) guarantees for video streaming over the Internet. Thus, we have to take congestion control and error control into consideration in order to meet the requirements mentioned above. The purpose of congestion control is to prevent packet loss and the goal of error control is to recover packet loss. In order to achieve congestion control, we adopt a new video compression framework called Fine Granular Scalability(FGS) which is capable of supporting a wide range of bandwidth-variation scenarios and being resilient to packet loss. Besides, since IP multicast is designed for efficiency and scalability, we utilize this technology as a means to transport video data. However, error control still needs to be concerned for IP multicast. In this paper, we present a reliable multicast video streaming system—FGS Reliable Multicast Applying FEC(Forward Error Correction). The characteristics of the system include: (1) a novel application-level protocol which can support loss recovery, (2) a buffer architecture which can access the video data quickly and help loss recovery, (3) some reliable mechanisms which integrate FEC and retransmission to achieve the saving of loss recovery time and network bandwidth usage. Finally, via experiments, we demonstrate the performance of the proposed system to show that it indeed provides high reliability in real-time video delivery.