Resilient wireless transmission of H.264/AVC through error localisation and control mechanisms

Abstract

Current trends in wireless communications provide for fast and location independent access to multimedia services. Due to its high compression efficiency, H.264/AVC is expected to become the dominant underlying technology in the delivery of future wireless video applications. However, H.264/AVC is susceptible to transmission errors common in wireless environments where even a single corrupted bit may cause visual artefacts that propagate in the spatio-temporal domain. The standard incorporates several error resilient mechanisms to minimize the effect of transmission errors on the perceptual quality of the reconstructed video sequence. However, these mechanisms assume a packet-loss scenario where all macroblocks (MBs) contained within a corrupted slice, including numerous uncorrupted MBs, are discarded and concealed. This implies that the error resilient mechanisms operate at a lower bound and thus further performance gains can be achieved by exploiting the residual redundancies available at the decoder side. During this dissertation, decoder-based techniques aimed to enhance the quality of damaged video sequences were investigated. The first method considered in this work exploits the residual source redundancy left by the standard encoder after compression to derive the most likelihood H.264/AVC feasible bitstream. This method manages to completely recover an average of 30% of the corrupted slices at no additional cost in bandwidth. The second approach considered in this dissertation exploits the redundancy available at pixel level to detect and localise visually distorted regions within the damaged slice that would otherwise be discarded. The experimental results show that machine learning algorithms can be taught to automatically detect the regions affected by transmission errors. This method limits the area to be concealed since only visually impaired regions are concealed. Both these methods provide a significant gain in video quality when compared to the standard when adopted individually. The two methods were combined together in a single solution to form the Hybrid Error Control Artefact Detection (HECAD) method which further boosts the performance of the individual components. This gain in performance is achieved at no additional cost in bandwidth and a moderate increase in complexity of the decoder. Furthermore, this method can be applied in conjunction with other error resilient strategies adopted by the standard decoder and still register considerable performance gains.