13.4.2007 HIIT seminar

HIIT seminars in spring 2007 will be held in hall **B222** of Exactum, on Fridays starting at 10:15 a.m. Coffee available from 10.

Fri Apr 13
Teemu Roos
Complexity, information, and noise: Denoising signals by the MDL principle
Denoising means the process of removing noise from a signal. This may be  necessary due to an imprecise measurement device or transmission over a  noisy channel. Traditional filtering techniques suppress the high-frequency components of the signal. This often removes a large fraction of the noise, but in some cases leads to loss of too much detail. Time-frequency transforms, including wavelet transforms, enable better resolution by operating both in the frequency domain and the time (spatial) domain.

We discuss the minimum description length (MDL) principle and its  foundations, in particular Kolmogorov's structure function. The MDL principle is by its very purpose designed to separate information and noise, and hence naturally applicable to denoising. We analyze an extend an earlier MDL denoising method of Rissanen.

(joint work with P. Myllymäki and J. Rissanen)

Last updated on 14 May 2007 by Martti Mäntylä - Page created on 13 Apr 2007 by Teija Kujala