[AstroPy] Interested in a wavelet toolbox?

Sat Apr 4 08:57:44 EDT 2015

On Fri, Apr 3, 2015 at 12:33 AM, Christoph Deil <
deil.christoph at googlemail.com> wrote:

>
> On 02 Apr 2015, at 23:40, Ricky Egeland <ricky.egeland at gmail.com> wrote:
>
> I was searching for a well-tested wavelet library in Python a few months
> ago, and did not find anything fully satisfactory.  I wanted a continuous
> wavelet transform using the Morlet mother wavelet, with significance
> contours.  I feel like such a package belongs in scipy more than it does
> astropy, as wavelets are a general signal processing technique not
> restricted to astronomy.
>
>
Hi,

Hi Amara,
>
> I agree with Ricky here … a good wavelets toolbox belongs in Scipy because
> it’s useful to many scientist and engineers.
> If you can contribute there, here’s a starting point:
> https://github.com/scipy/scipy/issues/3931 (and I’ve put Ralf Gommers as
> lead on this effort in CC).
>

A few comments:
- The main repo for that is at https://github.com/rgommers/pywt
- That code is a PyWavelets fork, so only DWT
- Quite a lot of effort went into it with contributions from 4 people. It's
close to being mergeable into Scipy, but it's currently stalled mainly
because I have very little bandwidth for it.
- Help is very much welcome. In particular, issues 46, 51 and 54 need
attention, plus review of and critical feedback on PR 55 would be very
useful.
- There's also interest from scikit-image for DWT:
https://github.com/scikit-image/scikit-image/pull/760

For continuous wavelets there has not been much recent work that I'm aware
of, and there is no decision or consensus over what code could be adapted
for use in Scipy. It is clear though that there's room in Scipy for
continuous wavelets, and that when a good feature-complete implementation
gets proposed it can go into scipy.wavelets. The current
scipy.signal.wavelets can then be deprecated.

Cheers,
Ralf

Then there’s probably going to be astronomy-specific wavelets and wavelet
> methods and these could go in Astropy or an Astropy-affiliated package.
> But I don’t think we should duplicate the functionality that’s already
> implemented in the packages listed below and that will hopefully soon
> become available in Scipy.
>
> Christoph
>
>  There are a few wavelet-related functions in scipy.signal already, but
> they do not work together (e.g. scipy.signal.cwt is unrelated to
> scipy.signal.morlet) and it is not clear how they are intended to be used.
> I have seen scipy discussions that propose throwing all that away and
> starting over.
>
> In case they are useful, here are my notes from the search:
>
> ===
> - scipy.signal.wavelet : cwt(), only Mexican hat wavelet?  Poor
> documentation
>
> - kPyWavelet : https://github.com/Cadair/kPyWavelet/
>   Linked from http://ocgweb.marine.usf.edu/~liu/wavelet.html
>   Few users... real interest?
>
> - A Practical Guide to Wavelet Analysis
>   http://paos.colorado.edu/research/wavelets/
>   Highly cited ... no python
>
> - pycwt
>   https://github.com/regeirk/pycwt
>   Explicitly based on T&C code above
>   # Tested.  Example code has a few minor bugs, but appears to be doing
> the right thing
>
> - PyCWT
>   https://github.com/Unidata/pyCWT
>   Simple, 1 file
>   Based on T&C
>   # Tested. No example code, but docstrings have examples (that don't work
> as-is)  Source code is cleaner than pycwt.  scaleogram() has some issues.
> Need to input scales, not frequencies...
>
> - ObsPy
>
> http://docs.obspy.org/master/tutorial/code_snippets/continuous_wavelet_transform.html
>
> - PyWavelets
>   http://www.pybytes.com/pywavelets/
>   Part of MacPorts
>   Good documentation effort
>   Only one developer, but 5+ years effort
>
> - https://github.com/aaren/wavelets
>
> - https://pypi.python.org/pypi/waipy
>   Based on T&C code
>
> - scipy wavelet discussions:
> https://github.com/scipy/scipy/issues?utf8=✓&q=is%3Aissue+is%3Aopen+wavelet
> ===
>
> In the end I forked Unidata/pyCWT, fixed a few things and used that (
> https://github.com/rickyegeland/pyCWT).  On a second look I decided
> regeirk/pycwt might have been the better implementation, but the interface
> and documentation were not the best for getting started.
>
> Hope this helps.
>
> Regards,
> Ricky Egeland
>
>
> On Apr 2, 2015, at 12:47 PM, Amara Graps <amara at konteur.com> wrote:
>
> Greetings astropy-ers,
>
> I'm proposing the development of a wavelet library in Python to a
> (European Commission) funding agency and I'll need beta testing
> of some parts. I think it would be especially great to have the
> library be incorporated into AstroPy after it's written and tested
> because the funds for maintenance would end when the project ends..
> yet astronomers continue to love wavelets.
>
> Some large part of my Python wavelet library would be built upon this
> well-established (general purpose) Matlab toolbox:
>
> http://statweb.stanford.edu/~wavelab/Wavelab_850/index_wavelab850.html
>
> Does this sound useful and interesting to this group? If it does, please
> reply to graps at psi.edu. I'm gathering documentation of the AstroPy user
> community's interest and my testing and maintenance methodology in order
> to help the project get funded.
>
> Thanks very much in advance!
>
> Sincerely,
> Amara
>
> Amara Graps, PhD
> Lead Researcher,  University of Latvia
> Astronomijas institūts, Raiņa bulvāris 19, Rīga, LV-1050, Latvia
> amara at lu.lv  +371 / 28853907
> and
> Senior Scientist, Planetary Science Institute (PSI, USA),
> Aleksandra Caka iela, 96-31,
> Riga, Latvia LV-1011
> http://www.psi.edu/about/staff/graps/graps.html
> graps at psi.edu  +1 / 644-6250
> _______________________________________________
> AstroPy mailing list
> AstroPy at scipy.org
> http://mail.scipy.org/mailman/listinfo/astropy
>
>
> _______________________________________________
> AstroPy mailing list
> AstroPy at scipy.org
> http://mail.scipy.org/mailman/listinfo/astropy
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/astropy/attachments/20150404/282b37a4/attachment.html>