help-octave
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Discrete wavelet analysis


From: peter
Subject: Re: Discrete wavelet analysis
Date: Mon, 30 Jun 2014 13:45:54 +0200 (CEST)
User-agent: SquirrelMail/1.4.8-21.el5.centos

Hi Helios,

> Hi,
>
> I'm interested in multiresolution analysis with wavelets in Octave. My
> starting point is some routines that use functions from Matlab's Wavelet
> Toolbox, so things would be easier if there were some package for Octave
> that defines them.
>
> However, I'm not sure what is the current situation in relation to this
> topic. I have read recommendations to using WaveLab, or the package
> ltfat, that implement a wide array of functions to work with wavelets.
> That might be ok for my purposes, although the transition from Matlab
> would not be that straightforward.
>
> I have failed to understand if the Octave community generally sees such
> packages as something like a "definitive" alternative to Matlab's
> wavelet toolbox, or it is just that a more "clonic" alternative is still
> not available, and in the meanwhile they are the best/more convenient
> tools to work with.
>
> I guess it may be the latter case, since the signal package contains
> the functions "dwt", "idwt" with identical definition as the ones of
> Matlab's package, although they are not still fully functional, since
> the function "wfilters" called by them is missing, at least until Octave
> 3.8. I'm also aware of the project for a wavelet package here, with
> "wfilters" and other functions:
> http://sourceforge.net/p/octave/wavelet/ci/default/tree/
> but it's not still published in Octave Forge.
>
> All in all, I would like to know what is the situation regarding this.
> Depending on the skills and background needed to complete that work, I
> would be willing to contribute to it, if there is work to be assigned.
>
> If this question is already solved elsewhere, or should be moved to
> octave-maintainers or wherever, please tell me.

I am the original author of LTFAT, so my view is perhaps a little biased,
but here goes:

We have intentionally not chosen to make a functionally equivalent version
of the Matlab Wavelet package, mainly because we never considered it to be
very good.  The output of the functions depends strongly on a global
variable, which is not a very portable behaviour.

In LTFAT there are very good implementations of the discrete wavelet
transform:

http://ltfat.sourceforge.net/doc/wavelets/fwt.php
http://ltfat.sourceforge.net/doc/wavelets/ifwt.php

and a bunch of different filters to choose from, see the "Filters defined
in the time-domain" headline here:

http://ltfat.sourceforge.net/doc/wavelets/index.php

Wavelab is also a good package, it has just not been updated for many
years, and as I recall you need to do some simple modification to get it
working in Octave. LTFAT works out of the box.

Cheers,
Peter.





reply via email to

[Prev in Thread] Current Thread [Next in Thread]