A forward error recovery technique for MPEG-II video transport

Abstract

Digital video is increasingly being used as an important medium of information. MPEG-II is widely used for compression of digital video in order to reduce the network bandwidth and storage requirements [4]. Digital video, when used in networked multimedia applications, suffers from data losses/errors. This is a serious problem in the case of wireless networks [7]. There are several ways to recover from these losses or errors. Recovery mechanisms based on re-transmission of the data may not be suitable in many cases because of the real-time nature of the applications and the absence of reverse channel for feedback. Real-time communication of digital video, as in the case of video conferencing, benefit from forward error correction/recovery techniques. Forward error correction (FEC) codes and frequent synchronizing codewords have been proposed in [5, 61. These techniques require low complexity hardware, however they add redundancy thereby lowering the coding efficiency. FEC can potentially add more errors to a bit stream once the correcting capability of the code has been exceeded. An error-resilient entropy coding (EREC) technique has been suggested in [2, 3, 11. EREC achieves bit stream resynchronization. [2, 3, l] also propose a hierarchical pyramid predictor in place of the standard differentially coded DPCM. These proposals consider channels subject to burst errors and packet errors and aim at providing an environment where performance degrades gracefully with increasing channel noise. Several simple error resilient approaches are suggested in [7]. These approaches include error concealment, temporal localization by sending extra I-frames, and spatial localization by providing early resynchronization of the elements in the bit stream that are coded differentially between MPEG-II macro-blocks.