PyImfit is a Python wrapper for Imfit, a C++-based program for fitting 2D models to scientific images. It is specialized for fitting astronomical images of galaxies, but can in principle be used to fit any 2D Numpy array of data.

WARNING: This is currently a work in progress; use at your own risk!

Changelog

Assuming you want to fit an astronomical image named galaxy.fits using a model defined in an Imfit configuration file named config_galaxy.dat:

from astropy.io import fits
import pyimfit

imageFile = "<path-to-FITS-file-directory>/galaxy.fits"
imfitConfigFile = "<path-to-config-file-directory>/config_galaxy.dat"

# read in image data, convert to proper double-precision, little-endian format
image_data = pyimfit.FixImage(fits.getdata(imageFile))

# construct model from config file (this can also be done programmatically within Python)
model_desc = pyimfit.ModelDescription.load(imfitConfigFile)

# create an Imfit object, using the previously created model configuration
imfit_fitter = pyimfit.Imfit(model_desc)

# load the image data and image characteristics (the specific values are
# for a typical SDSS r-band image, where a sky-background value of 130.14
# has already been subtracted), and then do a standard fit
# (using default chi^2 statistics and Levenberg-Marquardt solver)
imfit_fitter.fit(image_data, gain=4.725, read_noise=4.3, original_sky=130.14)

# check the fit and print the resulting best-fit parameter values
if imfit_fitter.fitConverged is True:
    print("Fit converged: chi^2 = {0}, reduced chi^2 = {1}".format(imfit_fitter.fitStatistic,
        imfit_fitter.reducedFitStatistic))
    bestfit_params = imfit_fitter.getRawParameters()
    print("Best-fit parameter values:", bestfit_params)

See the Jupyter notebook pyfit_emcee.ipynb in the docs subdirectory for an example of using PyImfit with the Markov Chain Monte Carlo code emcee.

Online documentation: https://pyimfit.readthedocs.io/en/latest/.

Also useful: Onine documentation for Imfit; and the main Imfit manual in PDF format: imfit_howto.pdf

PyImfit is designed to work with modern versions of Python 3 (3.5 or later on Linux; 3.6 or later on macOS); no support for Python 2 is planned.

A precompiled binary version ("wheel") of PyImfit for macOS can be installed from PyPI via pip:

$ pip3 install pyimfit

PyImfit requires the following Python libraries/packages (which will automatically be installed by pip if they are not already present):

• Numpy
• Scipy

Astropy is also useful for reading in FITS files as numpy arrays (and is required by the unit tests).

PyImfit can also be installed on Linux using pip. Since this involves building from source, you will need to have a working C++-11-compatible compiler (e.g., GCC version 4.8.1 or later); this is probably true for any reasonably modern Linux installation. (Note: a 64-bit Linux system is required.)

$ pip3 install pyimfit   [or "pip3 install --user pyimfit", if installing for your own use]

To build PyImfit from the Github source, you will need the following:

• Most of the same external (C/C++) libraries that Imfit requires: specifically FFTW3 [version 3], GNU Scientific Library [version 2.0 or later!], and NLopt

• This Github repository (use --recurse-submodules to ensure the Imfit repo is also downloaded)

    $ git clone --recurse-submodules git://github.com/perwin/pyimfit.git
  

• A reasonably modern C++ compiler -- e.g., GCC 4.8.1 or later, or any C++-11-aware version of Clang++/LLVM that includes support for OpenMP. See below for special notes about using the Apple-built version of Clang++ that comes with Xcode for macOS.

1. Install necessary external libraries (FFTW3, GSL, NLopt)

   • These can be installed from source, or via package managers (e.g., Homebrew on macOS)

   • Note that version 2.0 or later of GSL is required! (For Ubuntu, this means the libgsl-dev package for Ubuntu 16.04 or later.)

2. Clone the PyImfit repository

   $ git clone --recurse-submodules git://github.com/perwin/pyimfit.git
   

3. Build the Python package

   • [macOS only:] First, specify a valid, OpenMP-compatible C++ compiler

     $ export CC=<c++-compiler-name>; export CXX=<c++-compiler-name>
     

   (Note that you need to point CC and CXX to the same, Open-MP-compatible C++ compiler! This should not be necessary on a Linux system, assuming the default compiler is standard GCC.)

   • Versions of Apple's Clang compiler from Xcode 9 or later can compile OpenMP code, but you will need to also install the OpenMP library (e.g., brew install libomp if using Homebrew). See here for more details.

   • Build and install PyImfit!

     • For testing purposes (installs a link to current directory in your usual package-install location)

       $ python3 setup.py develop
       

     • For general installation (i.e., actually installs the package in your usual package-install location)

       $ python3 setup.py install
       

PyImfit originated as python-imfit, written by André Luiz de Amorim; the current, updated version is by Peter Erwin.

(See the Imfit manual for additional credits.)

PyImfit is licensed under version 3 of the GNU Public License.