Frame loss error concealment for spatial scalability using hallucination

We present a new error concealment algorithm for spatially scalable video coding with frame loss in the enhancement layer, based on the technique of hallucination. For a lost enhancement layer frame, the error concealment is done as hallucinating its base layer frame, using the database trained from previously decoded frames nearby to the lost one. Simulation results show that the proposed method could out-perform the state-of-the-art error concealment algorithms of SVC significantly.