Source code for psfsubtraction.fitpsf.fit

# Licensed under a MIT licence - see file `license`
import numpy as np


__all__ = ['psf_from_projection']


def psf_from_projection(self, image1d, psfbase):
    '''solve a linear algebra system for the best PSF

    Parameters
    ----------
    image1d : array in 1 dim
    psfbase : array in [M,N]
        M = number of pixels in flattened image
        N = number of images that form the space of potential PSFs

    Returns
    -------
    psf_coeff : array in 1 dim
        Coefficients for a linear combination of ``psfbase`` elements that
        that give the optimal PSF.

    Raises
    ------
    ValueError : If given masked data, because `numpy.linalg.solve` would silently
        use the "values behind the mask".
    '''
    if (np.ma.getmask(image1d).sum() > 0 or np.ma.getmask(psfbase).sum() > 0):
            raise ValueError('This fitter cannot deal with masked data.')

    a = np.dot(psfbase.T, psfbase)
    b = np.dot(psfbase.T, image1d)
    psf_coeff = np.linalg.solve(a, b)
    return psf_coeff