Error correction for MPEG-4 video transmission using VLECC

Abstract

When transmitting compressed video over data networks one has to deal with how channel errors affect the decoding process. This is particularly problematic with data loss or erasures. In this thesis we address the problem by using Variable Length Error Correcting Codes (VLECC) and sequential decoding using the stack algorithm on VLECC and Reversible Variable Length Codes (RVLC) at the decoder side. Currently MPEG4 uses RVLC in its standard implementation. RVLCs are special VLCs that can be uniquely decoded in both the forward and reverse directions. VLECC are VLC with additional structure to perform error correction. VLECC can also be made reversible hence inheriting the good properties of R VLC. The implementation has been based on the ISO MPEG4 codec (MoMuSys). Three types of code books have been analyzed in the project and statistics have been presented. The implementation modifies the original decoding strategies. The algorithm aims to correct errors on the video stream and hence improving video quality. Basic implementation details have also been outlined. Experiments have been conducted under various channel conditions (using an A WGN channel with E/N0 from 5 dB to 8.4 dB) and for two video sequences ("Stefan" and "Foreman") to demonstrate the effectiveness of the proposed scheme. On average an improvement in video quality measured in Peak Signal to Noise ratio (PSNR) of 8dB was noted using hard sequential decoding and 11 db using soft sequential (at a SNR varying from 6 dB to 8.6 dB) decoding at the expense of increased Macro Block decoding time of about ten times relative to the time taken by the standard MPEG4 decoding algorithm.