Frame based error concealment in H.264/AVC by refined motion prediction

In the transmission of low bit-rate video bit-stream, a packet loss usually results in whole frame loss, and the frame based error concealment (FBEC) becomes a necessary technique for video decoders in many real time applications. In this paper, we propose a method of FBEC to improve the visual quality in H.264/AVC decoders. The lost frames are reconstructed by predicting motion vectors from neighbor macro-blocks of the previous frames. With the proposed algorithm, the decoding performance measured in PSNR is around 1dB higher than that of JM10.0. The subjective quality also improves significantly. This method has little overhead of extra computing components, and may be easily implemented in hardware.