DocumentCode :
2328073
Title :
MMC02-5: A Low-Complexity Methodology for Unequal Error Protection of Scalable Images
Author :
Salemi, Eric ; Desset, Claude ; Dejonghe, Antoine ; Cornelis, Jan ; Schelkens, Peter
Author_Institution :
Interuniversity Microelectron. Center, Leuven
fYear :
2006
fDate :
Nov. 27 2006-Dec. 1 2006
Firstpage :
1
Lastpage :
5
Abstract :
Unequal error protection (UEP) is the most practical way to achieve joint source-channel coding while keeping a clean separation between the source and channel coders. In this paper, a generic UEP allocation methodology is proposed. The methodology is based on a run-time algorithm, exploiting design-time models, and is applicable to bit-level transmission scenarios as well as packet-based transmission environments. It works with source-independent modeling and a linear complexity with respect to the number of substreams in the source coded stream and the number of protection levels available, which makes it an attractive solution for practical scenarios. We obtain a substantial complexity reduction compared to a classical Lagrangian optimization while keeping optimality in the allocation of the protection levels. Given a typical source and channel granularity, we obtain a factor 6 complexity reduction.
Keywords :
combined source-channel coding; error statistics; bit-level transmission; classical Lagrangian optimization; exploiting design-time model; joint source-channel coding; packet-based transmission; run-time algorithm; scalable images; source-independent modeling; unequal error protection; Algorithm design and analysis; Bit rate; Channel coding; Error correction codes; Lagrangian functions; Microelectronics; Protection; Quantization; Resource management; Runtime environment;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Global Telecommunications Conference, 2006. GLOBECOM '06. IEEE
Conference_Location :
San Francisco, CA
ISSN :
1930-529X
Print_ISBN :
1-4244-0356-1
Electronic_ISBN :
1930-529X
Type :
conf
DOI :
10.1109/GLOCOM.2006.207
Filename :
4150837
Link To Document :
بازگشت