Python and Wavelets

[Ralf Muschall]

Phil Austin wrote:

> > here: http://www.eee.metu.edu.tr/~ngencer/amara.htm

She actualized her list from time to time, but www.amara.com
seems to be down since a few weeks, so one would have to search
the web (IIRC some site in .de with "mpi" in it).

> Two other good possibilities for wrapping:
> http://www-stat.stanford.edu/~wavelab/
> http://wave.cmap.polytechnique.fr/soft/LastWave.html

I looked at various places when I needed wavelets recently
and ended with downloading the coefficients from
http://phase.etl.go.jp/contrib/wavelet/
and hacking the C++ myself. But I have no idea how to
incorporate this into NumPy (or how to use the non-Daubechies
stuff).

The mentioned site has the coefficients up to rather
high orders with many decimal places and includes test
protocols which seem to indicate that they are correct
(the doc is in japanese, so I can only guess).
This important -- computing them is nontrivial, and
Daubechies' "10 lectures" is reported to have bugs.

I started from the description (not the code) in "Numerical
recipes". The code there halves the interval always down
until length==4, which seems strange to me -- I stopped when
it became as short as the wavelet.

Ralf