DocumentCode :
2932066
Title :
Loss-distortion estimation for practical H.264/AVC streams
Author :
Korhonen, Jari ; Perkis, Andrew
Author_Institution :
Centre for Quantifiable Quality of Service in Commun. Syst. (Q2S), Norwegian Univ. of Sci. & Technol. (NTNU), Trondheim, Norway
fYear :
2009
fDate :
June 28 2009-July 3 2009
Firstpage :
766
Lastpage :
769
Abstract :
Even though several unequal erasure protection (UEP) schemes have been proposed for video streaming, it is still a challenge to define the optimal relative protection level for different data units. In this paper, we study loss-distortion modeling in realistic video streaming scenarios. The main observation is that when video streams with complex hierarchical structures and error resilience features are considered, the relative significance of different units cannot be reliably estimated from mutual dependencies between units as suggested in related research. Therefore, data units can be classified reliably into different perceptual priority classes via analysis by synthesis only. For estimating the co-impact of multiple losses, we propose a simple loss-distortion model. The model can be used for optimizing UEP in practical streaming systems.
Keywords :
telecommunication security; video coding; video streaming; H.264-AVC streams; analysis by synthesis; error resilience features; loss-distortion estimation; optimal relative protection level; perceptual priority classes; unequal erasure protection schemes; video streaming; Automatic voltage control; Councils; Distortion measurement; Error correction codes; Forward error correction; PSNR; Quality of service; Resilience; Signal synthesis; Streaming media; Multimedia communication; Video quality; Video signal processing;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Multimedia and Expo, 2009. ICME 2009. IEEE International Conference on
Conference_Location :
New York, NY
ISSN :
1945-7871
Print_ISBN :
978-1-4244-4290-4
Electronic_ISBN :
1945-7871
Type :
conf
DOI :
10.1109/ICME.2009.5202607
Filename :
5202607
Link To Document :
بازگشت