Reference frame modification techniques for temporal and spatial scalability

摘要

Compressed video is usually transmitted over channels which are not necessarily error free. Channel errors can result in a mismatch between the encoder and the decoder, and because of the predictive structures used in video coding, the errors will propagate both temporally and spatially. Consequently, the quality of the received video at the decoder may degrade significantly. In order to improve the quality of the received video, several error resilient methods have been proposed. In this paper, the introduced mismatch is reduced by modifying the prediction structure and forming a more robust reference frame. The proposed technique combines error robustness of previous Intra coded blocks, better prediction achieved using the previous reference frame, and exponential decay of error propagation caused by the leaky prediction. This technique was examined with the scalable extension of H.264/AVC. Furthermore, the technique was also used in combination with random Intra refresh and error resilience mode decision techniques to achieve better robustness. Simulation results show the effectiveness of our scheme, especially for medium and high motion sequences.