Talk:Stationary wavelet transform

Latest comment: 15 years ago by Goossensb in topic The Name of the article

What is it? edit

Could someone please shed some light on what the use of this transform is, as opposed to a downsampling scheme? Thanks. 65.183.135.231 (talk) 16:50, 22 July 2008 (UTC)Reply

The Name of the article edit

The name "Stationary Wavelet transform" was coined by Nason and Silverman in ["The stationary wavelet transform and some statistical applications", in A. Antoniadis and G. Oppenheim, eds. Wavelets and statistics, volume 103 of Lecture notes in statistics, pp 281-229, Springer-Verlag, New York]. Since then, Nason has gone on denounce this name because it's misleading. In his 2008 book "Wavelet Methods in Statistics with R" page 60, he states "(Stationary Wavelet Transform) turns out not to be a good name because the non-decimated wavelet transform (the name he now uses) is actually useful for studying nonstationary time series. However, some older works occassionally refer to the older name".

Given that "Stationary wavelet transform" is a misleading name, and that it seems to have been supplanted by "Non-decimated wavelet transform" or "Maximal overlap wavelet transform" in the literature, I propose that the name of the article be changed. Unfortunately, Matlab seems to still be using SWT to refer to this transform. At the very least, all other names need to redirect here. 203.167.251.186 (talk) 02:30, 1 December 2008 (UTC) DRBReply

Does anyone know how to change the name of the article? Also, I think it would be instructive to carry out a pen and paper calculation of MODWT so readers can see the basic idea. Anyone up for it? 203.97.49.105 (talk) 07:20, 14 January 2009 (UTC)Reply

Registered users can move an article to a new name. But I'll wait for others to comment. For the moment, I will make sure that redirects from your proposed names exist. The maximmal overlap variant already existed.--LutzL (talk) 09:53, 14 January 2009 (UTC)Reply
In Image Processing literature (IEEE Trans. Image Proc.) the name "undecimated wavelet transform" is most commonly used to make a distinction with the decimated wavelet transform and sometimes (but not often) translation invariant WT is also used. Some names that are mentioned refer to a specific technique/algorithm (à trous, cycle spinning), others are often misleading, e.g. redundant wavelet transform: there are multiple ways to have a redundant WT (e.g. complex wavelets...). MODWT: this is the first time I hear this name, don't know where it comes from. So I would go for "undecimated wavelet transform" Goossensb (talk) 20:36, 4 March 2009 (UTC)Reply