Spatio-temporal video error concealment with perceptually optimized mode selection

We proposed a spatio-temporal error concealment algorithm for video transmission in an error-prone environment. The proposed technique employs motion vector estimation, edge-preserving interpolation, and texture analysis/synthesis. It has two main advantages with respect to existing methods, namely: i) it aims at optimizing the visual quality of the restored video, and not only PSNR; and ii) it employs an automatic mode selection algorithm in order to decide, on a macroblock basis, whether to use the spatial restoration, the temporal one, or a combination thereof. The algorithm has been applied to H26L video, providing satisfactory performance over a large set of operating conditions.