Error Detection using Data Embedding for H.263 Compatible Video Coding

Abstract

Due to the use of the variable length coding, a single error in the H.263 video bit stream may cause serious error propagation and quality degradation. Many error recovery techniques have been applied to H.263 video coding. Most importantly, the efficiency of the error recovery relies on the capability of the error detection. In this thesis, we propose a parity-embedded error detection technique for the H.263 compatible video coding. The requisite parity-check information is embedded in the encoded bit stream, which could be easily extracted by the decoder for detecting the actual error occurrence locations. At the encoder, all parity-check codes for all macroblocks in the previous frame are embedded into the motion vectors and the quantized residual DCT coefficients of the current encoding frame. Then the decoder can effectively manipulate these embedded bits to enhance the error detection capability at the macroblock-layer. From the simulation results, we can find that the proposed strategy is able to detect errors with extra 10~30% accuracy than the inherent variable length code provides at the bit error rate 10-2¡ã10-4. Moreover, the computation complexity is considerably low and the additional cost is only one-frame coding delay and little bitrate increase.

Index Terms - Error Detection, error concealment, H.263, data hiding, watermarking