Decoder motion vector estimation for scalable video error concealment

摘要

We propose a technique to recover lost enhancement layer information in scalable video using the information from: (1) the current base layer; and (2) the previous base- and enhancement-layer together with a decoder motion vector estimation method. An average of 1 dB improvement is reported for the enhancement layer vs. existing techniques. An accelerated method is also demonstrated to make the concealment real-time.