Entire-frame error concealment in H.264 using motion tracking and adaptive block sizes

Since H.264 provides high compression efficiency, a coded frame has a small size and can be encapsulated into a single network packet. Therefore, the loss of a single packet results in the loss of an entire frame, and then the error propagates to successive frames. In this paper, we propose a novel entire-frame error concealment method for H.264. First, the proposed method extrapolates motion vectors of 8×8 blocks and 4×4 blocks from neighboring frames onto the lost frame by motion tracking. Then, an appropriate conceal size of 4×4 or 8×8 is selected according to the overlapping area. Finally, the extrapolated motion vectors are refined by searching the best-match motion vector among the neighboring motion vectors. Experimental results show that the proposed method outperforms the conventional methods. This method is also applicable to the loss of non-entire frame.