Error concealment for spatially Scalable Video Coding using hallucination

We propose a new error concealment method based on hallucination for Scalable Video Coding with spatial scalability. In this method, parts of the frames which lose the enhancement layer are up-sampled from base layer and ldquohallucinatedrdquo as concealment frames. The database for hallucination is generated from the high-resolution and low-resolution frame-pairs near the lost frames in the video sequence. The effectiveness of hallucination here lies in the similarity between the nearby frames in the video sequence. Experiments show that the proposed method has superior results over the state-of-the-art error concealment method for spatially Scalable Video Coding.