Scalable video coding techniques are used to facilitate the delivery of streamed video content over networks with varying channel characteristics and unpredictable quality-of-service (QoS) such as the Internet and wireless Local-Area-Networks (LANs). Rate control methods are typically used with scalability solutions to adapt the video to variations in bandwidth availability. In this thesis, we develop two rate-distortion (RD) based rate-control algorithms for scalable video coded using the new fine-grained scalability (FGS) technique incorporated in the MPEG-4 video-coding standard. First, we show that the distortion measures, which are based on a square error criterion, can be computed using a very simple approach in the transform (DCT) domain. Second, we describe the two RD algorithms and their implementations within the transform domain of the FGS MPEG-4 framework. Finally, we present the results of experiments conducted to test the performance of the RD-based algorithms against the existing rate-control method. We find, upon analysis of these results in the transform domain, that the proposed rate-control algorithms provide optimum (convex) RD curves for a wide range of bitrates and video sequences. Improvements shown in the RD curves are particularly salient over bitrate ranges that can be supported over emerging Internet access technologies (e.g., DSL, cable modems, and wireless LANs).
展开▼
