Section: New Results
Regularity-preserving signal denoising
Participant : Jacques Lévy Véhel.
In collaboration with A. Echelard.
We have proposed a new wavelet-based method for signal denoising, that allows one to recover the local Hölder regularity of the original signal under weak assumptions  . The algotihm is a modification of the well-known wavelet thresholding procedure, where "small" coefficients are not put to zero, but modified in a way governed by the behaviour of large scale coefficients. This will have applications in the frame of our Tandem project on the analysis of radar images.