getting_started_1.py 5.27 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
# 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/>.
#
# Copyright(C) 2013-2018 Max-Planck-Society
#
16
17
# NIFTy is being developed at the Max-Planck-Institut fuer Astrophysik.

Philipp Arras's avatar
Philipp Arras committed
18
19
20
###############################################################################
# Compute a Wiener filter solution with NIFTy
# Shows how measurement gaps are filled in
21
# 1D (set mode=0), 2D (mode=1), or on the sphere (mode=2)
Philipp Arras's avatar
Philipp Arras committed
22
###############################################################################
23

24
import numpy as np
25

Philipp Arras's avatar
Philipp Arras committed
26
27
import nifty5 as ift

28

Philipp Arras's avatar
Philipp Arras committed
29
30
def make_checkerboard_mask(position_space):
    # Checkerboard mask for 2D mode
31
32
33
    mask = np.ones(position_space.shape)
    for i in range(4):
        for j in range(4):
Philipp Arras's avatar
Philipp Arras committed
34
35
            if (i + j) % 2 == 0:
                mask[i*128//4:(i + 1)*128//4, j*128//4:(j + 1)*128//4] = 0
36
37
    return mask

Philipp Arras's avatar
Philipp Arras committed
38

39
def make_random_mask():
Philipp Arras's avatar
Philipp Arras committed
40
    # Random mask for spherical mode
41
    mask = ift.from_random('pm1', position_space)
Philipp Arras's avatar
Philipp Arras committed
42
    mask = (mask + 1)/2
Martin Reinecke's avatar
Martin Reinecke committed
43
    return mask.to_global_data()
44

Philipp Arras's avatar
Philipp Arras committed
45

Philipp Arras's avatar
Philipp Arras committed
46
def mask_to_nan(mask, field):
Philipp Arras's avatar
Philipp Arras committed
47
    # Set masked pixels to nan for plotting
Philipp Arras's avatar
Philipp Arras committed
48
49
50
51
52
    masked_data = field.local_data.copy()
    masked_data[mask.local_data == 0] = np.nan
    return ift.from_local_data(field.domain, masked_data)


53
if __name__ == '__main__':
Philipp Arras's avatar
Philipp Arras committed
54
    np.random.seed(42)
Philipp Arras's avatar
Philipp Arras committed
55

Philipp Arras's avatar
Philipp Arras committed
56
    # Choose space on which the signal field is defined
Martin Reinecke's avatar
Martin Reinecke committed
57
    mode = 1
Philipp Arras's avatar
Philipp Arras committed
58
    if mode == 0:
Philipp Arras's avatar
Philipp Arras committed
59
        # One-dimensional regular grid
Philipp Arras's avatar
Philipp Arras committed
60
61
62
        position_space = ift.RGSpace([1024])
        mask = np.ones(position_space.shape)
    elif mode == 1:
Philipp Arras's avatar
Philipp Arras committed
63
        # Two-dimensional regular grid with checkerboard mask
Philipp Arras's avatar
Philipp Arras committed
64
        position_space = ift.RGSpace([128, 128])
Philipp Arras's avatar
Philipp Arras committed
65
        mask = make_checkerboard_mask(position_space)
Philipp Arras's avatar
Philipp Arras committed
66
    else:
Philipp Arras's avatar
Philipp Arras committed
67
        # Sphere with half of its pixels randomly masked
Philipp Arras's avatar
Philipp Arras committed
68
69
        position_space = ift.HPSpace(128)
        mask = make_random_mask()
70

Philipp Arras's avatar
Philipp Arras committed
71
    # Specify harmonic space corresponding to signal
72
    harmonic_space = position_space.get_default_codomain()
73

Philipp Arras's avatar
Philipp Arras committed
74
75
    # Harmonic transform from harmonic space to position space
    HT = ift.HarmonicTransformOperator(harmonic_space, target=position_space)
76

Philipp Arras's avatar
Philipp Arras committed
77
78
    # Set prior correlation covariance with a power spectrum leading to
    # homogeneous and isotropic statistics
79
    def power_spectrum(k):
Philipp Arras's avatar
Philipp Arras committed
80
81
82
83
84
85
86
87
88
        return 100./(20. + k**3)

    # 1D spectral space on which the power spectrum is defined
    power_space = ift.PowerSpace(harmonic_space)

    # Mapping to (higher dimensional) harmonic space
    PD = ift.PowerDistributor(harmonic_space, power_space)

    # Apply the mapping
89
    prior_correlation_structure = PD(ift.PS_field(power_space, power_spectrum))
Philipp Arras's avatar
Philipp Arras committed
90
91

    # Insert the result into the diagonal of an harmonic space operator
92
    S = ift.DiagonalOperator(prior_correlation_structure)
Philipp Arras's avatar
Philipp Arras committed
93
    # S is the prior field covariance
94

Philipp Arras's avatar
Philipp Arras committed
95
    # Build instrument response consisting of a discretization, mask
96
    # and harmonic transformaion
97

Philipp Arras's avatar
Philipp Arras committed
98
    # Data lives in a geometry-free space, thus the geometry is removed
99
    GR = ift.GeometryRemover(position_space)
Philipp Arras's avatar
Philipp Arras committed
100
101

    # Masking operator to model that parts of the field have not been observed
102
    mask = ift.Field.from_global_data(position_space, mask)
103
    Mask = ift.DiagonalOperator(mask)
Philipp Arras's avatar
Philipp Arras committed
104
105
106

    # The response operator consists out of
    # - an harmonic transform (to get to image space)
107
108
    # - the application of the mask
    # - the removal of geometric information
Martin Reinecke's avatar
Martin Reinecke committed
109
    R = GR(Mask(HT))
110
111
112

    data_space = GR.target

113
    # Set the noise covariance N
114
115
    noise = 5.
    N = ift.ScalingOperator(noise, data_space)
116

Philipp Arras's avatar
Philipp Arras committed
117
    # Create mock data
118
119
120
    MOCK_SIGNAL = S.draw_sample()
    MOCK_NOISE = N.draw_sample()
    data = R(MOCK_SIGNAL) + MOCK_NOISE
121

Philipp Arras's avatar
Philipp Arras committed
122
    # Build inverse propagator D and information source j
Martin Reinecke's avatar
Martin Reinecke committed
123
    D_inv = R.adjoint(N.inverse(R)) + S.inverse
124
    j = R.adjoint_times(N.inverse_times(data))
Philipp Arras's avatar
Philipp Arras committed
125
    # Make D_inv invertible (via Conjugate Gradient)
126
    IC = ift.GradientNormController(iteration_limit=500, tol_abs_gradnorm=1e-3)
127
    D = ift.InversionEnabler(D_inv, IC, approximation=S.inverse).inverse
128

Philipp Arras's avatar
Philipp Arras committed
129
    # Calculate WIENER FILTER solution
130
131
    m = D(j)

Philipp Arras's avatar
Philipp Arras committed
132
    # Plotting
Philipp Arras's avatar
Philipp Arras committed
133
    rg = isinstance(position_space, ift.RGSpace)
134
    plot = ift.Plot()
Philipp Arras's avatar
Philipp Arras committed
135
    if rg and len(position_space.shape) == 1:
Philipp Arras's avatar
Philipp Arras committed
136
137
138
139
140
141
        plot.add(
            [HT(MOCK_SIGNAL), GR.adjoint(data),
             HT(m)],
            label=['Mock signal', 'Data', 'Reconstruction'],
            alpha=[1, .3, 1])
        plot.add(mask_to_nan(mask, HT(m - MOCK_SIGNAL)), title='Residuals')
142
        plot.output(nx=2, ny=1, xsize=10, ysize=4, title="getting_started_1")
Philipp Arras's avatar
Philipp Arras committed
143
    else:
144
        plot.add(HT(MOCK_SIGNAL), title='Mock Signal')
Philipp Arras's avatar
Philipp Arras committed
145
        plot.add(mask_to_nan(mask, (GR(Mask)).adjoint(data)), title='Data')
146
        plot.add(HT(m), title='Reconstruction')
Philipp Arras's avatar
Philipp Arras committed
147
        plot.add(mask_to_nan(mask, HT(m - MOCK_SIGNAL)), title='Residuals')
148
        plot.output(nx=2, ny=2, xsize=10, ysize=10, title="getting_started_1")