DocumentCode :
3049351
Title :
Universal lossless compression of piecewise stationary slowly varying sources
Author :
Shamir, G.I. ; Costello, Daniel J.
Author_Institution :
Dept. of Electr. Eng., Notre Dame Univ., IN, USA
fYear :
2001
fDate :
27-29 March 2001
Firstpage :
371
Lastpage :
380
Abstract :
Universal lossless compression of parametric piecewise stationary sources with slow changes in the statistics between stationary segments that take place in unknown time intervals is investigated. The minimum description length (MDL) principle is derived for two different settings of this problem under the assumption that the parameter changes are linear over the change interval. In the first setting, it is assumed that all changes are of equal known in advance duration d, and in the second setting all statistics changes are of unknown durations. While in both cases the redundancy for most sources for each unknown statistical parameter in each segment remains lower bounded, as in the case of abruptly changing statistics, by 0.5 log m extra code bits, where m is the mean segment length, the minimum extra code-length required for each unknown transition interval decreases to log m-0.5 log d in the first setting, but surprisingly remains log m, as in the case of abruptly changing statistics, in the second. Schemes that achieve the lower bounds in both settings are demonstrated.
Keywords :
data compression; redundancy; source coding; statistical analysis; MDL principle; known durations; lower bounds; minimum description length principle; minimum extra code-length; parametric piecewise stationary sources; slowly varying sources; statistics changes; transition interval; universal coding; universal lossless compression; unknown durations; unknown statistical parameter; Costs; Entropy; Gas insulated transmission lines; Image segmentation; Layout; NASA; Parametric statistics;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Data Compression Conference, 2001. Proceedings. DCC 2001.
Conference_Location :
Snowbird, UT, USA
ISSN :
1068-0314
Print_ISBN :
0-7695-1031-0
Type :
conf
DOI :
10.1109/DCC.2001.917168
Filename :
917168
Link To Document :
بازگشت