rgrgtransformation.py

# NIFTy
# Copyright (C) 2017  Theo Steininger
#
# Author: Theo Steininger
#
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program.  If not, see <http://www.gnu.org/licenses/>.

import numpy as np
from transformation import Transformation
from rg_transforms import FFTW, NUMPYFFT
from nifty import RGSpace, nifty_configuration


class RGRGTransformation(Transformation):
    def __init__(self, domain, codomain=None, module=None):
        super(RGRGTransformation, self).__init__(domain, codomain,
                                                 module=module)

        if module is None:
            if nifty_configuration['fft_module'] == 'fftw':
                self._transform = FFTW(self.domain, self.codomain)
            elif nifty_configuration['fft_module'] == 'numpy':
                self._transform = NUMPYFFT(self.domain, self.codomain)
            else:
                raise ValueError('ERROR: unknow default FFT module:' +
                                 nifty_configuration['fft_module'])
        else:
            if module == 'fftw':
                self._transform = FFTW(self.domain, self.codomain)
            elif module == 'numpy':
                self._transform = NUMPYFFT(self.domain, self.codomain)
            else:
                raise ValueError('ERROR: unknow FFT module:' + module)

    @classmethod
    def get_codomain(cls, domain, zerocenter=None):
        """
            Generates a compatible codomain to which transformations are
            reasonable, i.e.\  either a shifted grid or a Fourier conjugate
            grid.

            Parameters
            ----------
            domain: RGSpace
                Space for which a codomain is to be generated
            zerocenter : {bool, numpy.ndarray}, *optional*
                Whether or not the grid is zerocentered for each axis or not
                (default: None).

            Returns
            -------
            codomain : nifty.rg_space
                A compatible codomain.
        """
        if not isinstance(domain, RGSpace):
            raise TypeError("domain needs to be a RGSpace")

        # parse the zerocenter input
        if zerocenter is None:
            zerocenter = domain.zerocenter
        # if the input is something scalar, cast it to a boolean
        else:
            temp = np.empty_like(domain.zerocenter)
            temp[:] = zerocenter
            zerocenter = temp

        # calculate the initialization parameters
        distances = 1 / (np.array(domain.shape) *
                         np.array(domain.distances))

        new_space = RGSpace(domain.shape,
                            zerocenter=zerocenter,
                            distances=distances,
                            harmonic=(not domain.harmonic))

        # better safe than sorry
        cls.check_codomain(domain, new_space)
        return new_space

    @classmethod
    def check_codomain(cls, domain, codomain):
        if not isinstance(domain, RGSpace):
            raise TypeError("domain is not a RGSpace")

        if not isinstance(codomain, RGSpace):
            raise TypeError("domain is not a RGSpace")

        if not np.all(np.array(domain.shape) ==
                      np.array(codomain.shape)):
            raise AttributeError("The shapes of domain and codomain must be "
                                 "identical.")

        if domain.harmonic == codomain.harmonic:
            raise AttributeError("domain.harmonic and codomain.harmonic must "
                                 "not be the same.")

        # Check if the distances match, i.e. dist' = 1 / (num * dist)
        if not np.all(
            np.absolute(np.array(domain.shape) *
                        np.array(domain.distances) *
                        np.array(codomain.distances) - 1) <
                10**-7):
            raise AttributeError("The grid-distances of domain and codomain "
                                 "do not match.")

        super(RGRGTransformation, cls).check_codomain(domain, codomain)

    def transform(self, val, axes=None, **kwargs):
        """
        RG -> RG transform method.

        Parameters
        ----------
        val : np.ndarray or distributed_data_object
            The value array which is to be transformed

        axes : None or tuple
            The axes along which the transformation should take place

        """
        if self._transform.codomain.harmonic:
            # correct for forward fft.
            # naively one would set power to 0.5 here in order to
            # apply effectively a factor of 1/sqrt(N) to the field.
            # BUT: the pixel volumes of the domain and codomain are different.
            # Hence, in order to produce the same scalar product, power===1.
            val = self._transform.domain.weight(val, power=1, axes=axes)

        # Perform the transformation
        Tval = self._transform.transform(val, axes, **kwargs)

        if not self._transform.codomain.harmonic:
            # correct for inverse fft.
            # See discussion above.
            Tval = self._transform.codomain.weight(Tval, power=-1, axes=axes)

        return Tval