25 October 2004 Rate-distortion-based mode selection for H.264/AVC in wireless environments
Author Affiliations +
Proceedings Volume 5600, Multimedia Systems and Applications VII; (2004) https://doi.org/10.1117/12.570789
Event: Optics East, 2004, Philadelphia, Pennsylvania, United States
Abstract
In the most recent H.264/AVC video codec, a loss-aware rate-distortion based coding mode selection method uses K copies of random variable channel behavior and decoder pairs to estimate the expected distortion at the encoder. The expected distortion can be estimated accurately if K is chosen large enough. However, the added complexity at the encoder is obviously at least K times the decoder complexity. In this paper, we modify another popular rate-distortion based mode selection method, the recursive optimal pre-sample estimate (ROPE) algorithm, by recursively calculating the standard deviation of pixel samples instead of the second moment values. Since the standard deviation values have the same order as the first moment of samples, the revised ROPE algorithm can be easily adopted to the H.264/AVC encoding process. Our experimental results show that the H.264/AVC coder with the revised ROPE algorithm has comparable error resilient capabilities, but it has only between one third to one half of the computation complexity of the current H.264/AVC codec with K=30.
© (2004) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Yun Gong, "Rate-distortion-based mode selection for H.264/AVC in wireless environments", Proc. SPIE 5600, Multimedia Systems and Applications VII, (25 October 2004); doi: 10.1117/12.570789; https://doi.org/10.1117/12.570789
PROCEEDINGS
7 PAGES


SHARE
RELATED CONTENT

Adaptive MPEG-2 information structuring
Proceedings of SPIE (January 22 1999)
MPML-based error-resilient wireless video transmission
Proceedings of SPIE (November 19 2003)
The need for an Internet video compression standard
Proceedings of SPIE (September 24 2011)

Back to Top