Unequal Forced-Intra-Refresh for robust video streaming

INTRODUCTION In the encoding or transcoding of video stream- ing, intra refresh is introduced as a non-normative error-resilience tool. Although the intra-refresh schemes based on the rate-distortion optimization (C?té et al., 2000; Stockhammer et al., 2002) are superior to the early heuristic intra-refresh strategies, these new schemes cannot yet accurately estimate the video distortion due to time-varying packet losses, and usually ignore the inter-frame error-propagation effect. Chiou…

Unequal forced-intra-refresh for robust video streaming

For video streaming over lossy channels, intra refresh can mitigate the error-propagation effect caused by packet losses. Besides some intra-mode macroblocks (MBs) generated by the “Lagrangian rate-distortion” or “Sum of absolute difference” mode decision, the encoder or transcoder possibly needs to increase some “forced” intra-mode MBs for robust video streaming. Based on the error-propagation analysis in a group of pictures (GOP), we propose an unequal Forced-Intra-Refresh (FIR) scheme to improve packet loss resilience of video streaming. According to a GOP-level error-propagation model, the proposed unequal FIR scheme can optimally increase the unequal number of forced intra-mode MBs for different frames in a GOP. Simulation results showed that the proposed scheme can effectively enhance the robustness of video streaming under different channel conditions, and achieve about 0.1∼0.9 dB gains over the average FIR scheme in H.264/AVC tools. Project (No. 60332030) supported by the National Natural Science Foundation of China