Design and applications of MPEG video markup language (MPML).

摘要

The traditional video coding schemes follow a sequential processing framework, where some desirable video applications cannot be easily accomplished such as random access and error resilience. In this research, we devise a new XML-based MPEG Video Markup Language (MPML) to describe the MPEG video content. MPML presents a portable description of applications and provides important information existing in the video stream. With MPML, the video content can be easily published in the Web community, and its interoperability can be easily supported for long-term maintenance and storage.; By taking into account the special structure of MPML, we propose an efficient compression algorithm for MPML documents. Then, we investigate two MPML-based video coding applications. The first application is the error-resilient transmission over a noisy wireless environment. MPML can be customized to provide unequal error protection for MPEG contents against channel noise. With MPML, the macroblock-level error resilient technique is explored. The proposed MPML-based error resilience can result in better PSNR performance as compared with the state-of-the-art error resilient techniques. Synchronization is a challenging problem faced by time-critical video multicast applications in an unreliable network environment. As the second application of MPML, we present an MPML-based solution for the multicast synchronization problem, which can achieve the streaming resynchronization at the frame level in an error-prone environment.; Error corruption models (ECMs) are proposed to describe the interframe error propagation (EP) effect. Similar to a flowing fluid, the interframe EP can be characterized by the damping, shaping and drifting effects. In this research, we propose a fluid-based ECM (F-ECM) to characterize the EP effects in both spatial and temporal domains. By analyzing the motion compensated prediction scheme, we derive a diffusion equation and discuss its solution, which captures the damping effect. A tracking quadrilateral (TQ) algorithm and other algorithms are proposed to capture the shaping and drifting effects. Finally, we integrate all these algorithms to form an adaptive fluid-based ECM. We prove that the dependency weight approach used in the MPML-based error resilience is a direct application of the proposed F-ECM.