A model-based error concealment method for scalable video

Abstract

Scalable video compression algorithms based on motion-compensated spatio-temporal wavelet decomposition have recently gained a lot of interest for video transmission over packet networks due to their flexibility and coding performance. Meanwhile, the adaptation of scalable video to the variations of the transmission conditions over the network could result in losing a part of the bitstream. In this paper we show that a statistical spatio-temporal model of the wavelet coefficients can be effective in the prediction of the lost subbands at the decoder side. Based on this analysis we propose an error concealment method and provide experimental results showing the quality improvement that can be achieved by using it.