Transmission distortion analysis for H.264 video over wireless networks

Transmission distortion is the degradation of the video presentation quality incurred by the transmission error. It is a considerably challenging task to transmission distortion analysis, because the wireless channel is time-varying and error-prone. More importantly, the transmission errors in one video frame may propagate further due to temporal coding. In this paper, we develop a sequence-segment-level (SS-level) transmission distortion model for H.264 video streams. Unlike previous schemes, which were all based on the tempo-spatial domain and needed full decoding of the compressed bit stream, we propose a compressed domain approach to transmission distortion modeling. The model estimates the transmission distortion by combining the video features directly extracted from the compression domain with the characteristics of the wireless channel. Since the approach does not require full-decompression of the video stream, the proposed model has very low computational complexity, which is well suited for the real-time application. Experimental results show the proposed model has high accuracy.