Video Error Resilience through Efficient Shadowing of Decoder

A common method for stopping error propagation in encoded video is to transmit intracoded blocks whenever errors occur, but this results in significantly increased bitrate. Error propagation can also be stopped by synchronizing the state of the encoder to that of the decoder. For example, one may emulate and repeat the entire decoding – including the channel errors – inside the encoder. This approach has both a simple decoder as well as good rate-distortion performance, but has not been widely used because it is cumbersome at the encoder. In this paper we propose an efficient error shadowing technique that achieves encoder re-synchronization without the need for a full decoder-at-encoder implementation. Our encoder tracks the error magnitudes without explicitly recalculating the intensity values. Simulations reveal that the proposed method has better performance compared to other techniques that have comparable decoder complexity. The use of motion compensated prediction in the standard based video coders causes spatiotemporal error propagation at the decoder when the compressed bitstream is received erroneously. When limited feedback is available, various solutions to recover from error propagation have been proposed. They can be classified into three categories according to where extra complexity is introduced: encoder based [1, 2], decoder-based [3] and encoder-decoder methods like reference picture selection (RPS) in H.263+. Our method belongs to the class of encoder-based methods which are more attractive when the decoder must be kept simple. Motion Compensated Predictor Q Decoder +