Block-based weighted pixel-value superposition algorithm for inter-frame error concealment

Due to channel noise or congestion, video data packets can be lost when transmit in error-prone networks. In addition, compressed video sequences are very fragile to transmission errors. Most of inter-frame error concealment (EC) methods estimate one motion vector (MV) for a corrupted macroblock (MB) or sub-partition. This may result in part of the boundary of the reconstruction MB or subpartition discontinuous. To solve this problem, in this paper, we propose a block-based weighted pixel-value superposition (BWPS) algorithm for inter-frame EC, which estimates multiple MVs for a corrupted block, predicts the corrupted block using each MV, and then carries out the weighted pixel-value superposition to reconstruct the corrupted block. At the same time, our proposed EC method has a low computation complexity. We also propose a dynamic switch between BWPS algorithm and boundary matching algorithm (BMA) to improve the visual quality of the reconstructed signals. Simulation results show our proposed methods outperform the reference methods in H.264/AVC.