zindy / libatrous

My "A trous" wavelet library with Python bindings.

(edge aware) libatrous

My "A trous" wavelet library with Python 2 bindings.

This is a library I use to band-pass filter our microscopy images. For example, speckled noise and background fluorescence can both be removed by defining an adequate band (more to come).

There are a few (hopefully useful) tricks I have implemented in the library:

• The library uses separable kernels and OpenMP for speed improvements.
• Image border artifacts are greatly reduced by using mirrored edges.
• Rectangular shaped voxels are taken into account by using re-gridding (interpolated nearest neighbour positions).
• Edge awareness code. Allows the filtering of an image while preserving its edges.

Watch this space for additional code and tutorials. For now, see http://nbviewer.jupyter.org/github/zindy/libatrous/blob/master/GetStarted.ipynb

To build the Python library on 64 bit Windows with tdm-gcc, SWIG and Anaconda:

Things to download

• Ananconda Python. This was tested with 2.7
• SWIG (download swig-win)
• Cygwin
• tdm64-gcc

Package installation notes

• tdm64-gcc: When installing, don't forget to click "openmp" under gcc.
• SWIG: Make a note where you unzip the swigwin folder (e.g. a folder under C:\TDM-GCC-64)
• Anaconda: Global install, make sure %PYTHONPATH% is defined.
• Cygwin: A basic install will do, no need to install gcc

Additional steps

In a command window, type the following to create a libpython27.a usable by gcc:

"C:\TDM-GCC-64\x86_64-w64-mingw32\bin\gendef.exe" %PYTHONPATH%\python27.dll
"C:\TDM-GCC-64\bin\dlltool.exe" --as-flags=--64 -m i386:x86-64 -k -l libpython27.a -d python27.def
move libpython27.a C:\TDM-GCC-64\x86_64-w64-mingw32\lib
move "%PYTHONPATH%\libs\python27.lib" "%PYTHONPATH%\libs\python27.lib____"

You will need to modify the \Anaconda2\Lib\distutils\cygwincompiler.py as per this stackoverflow solution:

4. In the same module, modify get_msvcr() to return an empty list instead of ['msvcr90'] when msc_ver == '1500' .

Then launch cygwin terminal and add a line to your .bashrc to add the Anaconda Python, tdm64-gcc and SWIG paths:

export PATH=/cygdrive/c/Anaconda2:/cygdrive/c/TDM-GCC-64/swigwin-3.0.11:/cygdrive/c/TDM-GCC-64/bin:$PATH

Relaunch the cygwin terminal and these should all return values:

gcc -v
swig -version
python --version

In place compilation:

cd src
python setup.py build_ext --inplace

In place compilation in Linux:

Make sure you have python, gcc and swig installed. You will need to change the src/setup.cfg to

[build]
compiler=gcc

[build_ext]
inplace=1

Then it's the same

cd src
python setup.py build_ext --inplace

Finally...

Finally, you can copy the libatrous.py and _libatrous.pyd or _libatrous.so where you need them. For example, you can copy these in the tests folder and try test_libatrous.py

I think this covers about all the steps. If you have a simpler workflow, I would of course be happy to hear about it and amend these instructions.

libatrous's Issues

Image reconstruction using libatrous

Hello, I have been using libatrous to compute the à trous wavelet of some fluorescence microscopy images using the kernel B3 Spline 5x5m nos I would like to reconstruct my images using just some of the scales obtained but I can find a function or examples that do it. I wanted to use pywavelet but they do not have this kernel in their wavelet library. Do you know how I can apply the image reconstruction using my images?

Thank you
Xareni