CSC with a Spatial Mask¶
This example demonstrates the use of cbpdn.AddMaskSim for
convolutional sparse coding with a spatial mask
[50]. If the sporco-cuda extension is
installed and a GPU is available, a GPU accelerated version is used. The
example problem is inpainting of randomly distributed corruption of a
greyscale image. This is the same problem solved by example
cbpdn_ams_gry, but with a different approach to handling of the
lowpass image components. In this example, instead of pre-processing
with a non-linear lowpass filter, the lowpass components are represented
within the main optimisation problem via an impulse filter with gradient
regularization on the corresponding coefficient map (see Sec. 3 and Sec.
4 of [52]).
from __future__ import print_function
from builtins import input
import pyfftw # See https://github.com/pyFFTW/pyFFTW/issues/40
import numpy as np
from sporco import util
from sporco import signal
from sporco import metric
from sporco import plot
plot.config_notebook_plotting()
from sporco.admm import cbpdn
from sporco.fft import fftconv
from sporco import cuda
# If running in a notebook, try to use wurlitzer so that output from the CUDA
# code will be properly captured in the notebook.
sys_pipes = util.notebook_system_output()
Load a reference image.
img = util.ExampleImages().image('monarch.png', zoom=0.5, scaled=True,
gray=True, idxexp=np.s_[:, 160:672])
Create random mask and apply to reference image to obtain test image.
(The call to numpy.random.seed ensures that the pseudo-random noise
is reproducible.)
np.random.seed(12345)
frc = 0.5
msk = signal.rndmask(img.shape, frc, dtype=np.float32)
imgw = msk * img
Define pad and crop functions.
pn = 8
spad = lambda x: np.pad(x, pn, mode='symmetric')
zpad = lambda x: np.pad(x, pn, mode='constant')
crop = lambda x: x[pn:-pn, pn:-pn]
Construct padded mask and test image.
mskp = zpad(msk)
imgwp = spad(imgw)
Load dictionary.
D = util.convdicts()['G:8x8x128']
di = np.zeros(D.shape[0:2] + (1,), dtype=np.float32)
di[0, 0] = 1
Di = np.dstack((di, D))
Set up weights for the \(\ell_1\) norm to disable regularization of the coefficient map corresponding to the impulse filter intended to represent lowpass image components (not to be confused with the AMS impulse filter used to implement spatial masking).
wl1 = np.ones((1,)*2 + (Di.shape[2:]), dtype=np.float32)
wl1[..., 0] = 0.0
wl1i = np.concatenate((wl1, np.zeros(wl1.shape[0:-1] + (1,))), axis=-1)
When representing lowpass image components using an impulse filter
together with an \(\ell_2\) norm on the gradient of its coefficient
map, we usually want to set the weight array for this norm (specified by
the GradWeight option) to disable regularization of all coefficient
maps except for the one corresponding to that impulse filter (not to be
confused with the AMS impulse filter used to implement spatial masking).
In this case set a non-zero value for the weights of the other
coefficient maps size this improves performance in this inpainting
problem.
#wgr = np.zeros((Di.shape[2]), dtype=np.float32)
wgr = 2e-1 * np.ones((Di.shape[2]), dtype=np.float32)
wgr[0] = 1.0
wgri = np.hstack((wgr, np.zeros((1,))))
Set up admm.cbpdn.ConvBPDNGradReg options.
lmbda = 1e-2
mu = 2e-1
opt = cbpdn.ConvBPDNGradReg.Options({'Verbose': True, 'MaxMainIter': 200,
'HighMemSolve': True, 'RelStopTol': 5e-3,
'AuxVarObj': False, 'RelaxParam': 1.8,
'rho': 5e1*lmbda + 1e-1, 'L1Weight': wl1,
'GradWeight': wgr, 'AutoRho': {'Enabled': False,
'StdResiduals': False}})
Construct admm.cbpdn.AddMaskSim wrapper for
admm.cbpdn.ConvBPDNGradReg and solve via wrapper. If the
sporco-cuda extension is installed and a GPU is available, use the
CUDA implementation of this combination.
if cuda.device_count() > 0:
opt['L1Weight'] = wl1
opt['GradWeight'] = wgr
ams = None
print('%s GPU found: running CUDA solver' % cuda.device_name())
tm = util.Timer()
with sys_pipes(), util.ContextTimer(tm):
X = cuda.cbpdngrdmsk(Di, imgwp, mskp, lmbda, mu, opt)
t = tm.elapsed()
imgr = crop(np.sum(fftconv(Di, X, axes=(0, 1)), axis=-1))
else:
opt['L1Weight'] = wl1i
opt['GradWeight'] = wgri
ams = cbpdn.AddMaskSim(cbpdn.ConvBPDNGradReg, Di, imgwp, mskp, lmbda, mu,
opt=opt)
X = ams.solve().squeeze()
t = ams.timer.elapsed('solve')
imgr = crop(ams.reconstruct().squeeze())
GeForce RTX 2080 Ti GPU found: running CUDA solver
Itn Fnc DFid Regℓ1 Regℓ2∇ r s ρ
--------------------------------------------------------------------------
0 3.55e+02 1.60e+02 1.87e+04 3.89e+01 5.80e-01 1.48e+00 6.00e-01
1 2.80e+02 2.58e+01 2.34e+04 9.95e+01 3.81e-01 6.79e-01 6.00e-01
2 2.62e+02 2.97e+01 2.07e+04 1.27e+02 2.64e-01 9.15e-01 6.00e-01
3 1.74e+02 1.15e+01 1.44e+04 9.41e+01 1.35e-01 4.51e-01 6.00e-01
4 1.61e+02 4.48e+00 1.39e+04 8.77e+01 1.06e-01 2.41e-01 6.00e-01
5 1.41e+02 3.15e+00 1.20e+04 8.79e+01 7.19e-02 2.65e-01 6.00e-01
6 1.29e+02 3.39e+00 1.08e+04 8.37e+01 5.07e-02 1.96e-01 6.00e-01
7 1.13e+02 2.49e+00 9.54e+03 7.77e+01 4.22e-02 1.63e-01 6.00e-01
8 1.05e+02 1.99e+00 8.84e+03 7.39e+01 3.54e-02 1.56e-01 6.00e-01
9 9.92e+01 2.14e+00 8.29e+03 7.05e+01 3.05e-02 1.41e-01 6.00e-01
10 9.19e+01 2.05e+00 7.63e+03 6.77e+01 2.72e-02 1.37e-01 6.00e-01
11 8.95e+01 1.84e+00 7.46e+03 6.50e+01 2.51e-02 1.29e-01 6.00e-01
12 8.49e+01 1.81e+00 7.06e+03 6.24e+01 2.30e-02 1.25e-01 6.00e-01
13 8.02e+01 1.73e+00 6.65e+03 5.97e+01 2.14e-02 1.22e-01 6.00e-01
14 7.67e+01 1.66e+00 6.36e+03 5.73e+01 2.00e-02 1.18e-01 6.00e-01
15 7.39e+01 1.62e+00 6.12e+03 5.51e+01 1.89e-02 1.15e-01 6.00e-01
16 7.12e+01 1.56e+00 5.90e+03 5.31e+01 1.79e-02 1.12e-01 6.00e-01
17 6.82e+01 1.51e+00 5.64e+03 5.13e+01 1.69e-02 1.10e-01 6.00e-01
18 6.53e+01 1.47e+00 5.39e+03 4.94e+01 1.61e-02 1.09e-01 6.00e-01
19 6.31e+01 1.43e+00 5.21e+03 4.77e+01 1.54e-02 1.07e-01 6.00e-01
20 6.12e+01 1.38e+00 5.06e+03 4.61e+01 1.48e-02 1.05e-01 6.00e-01
21 5.91e+01 1.34e+00 4.88e+03 4.45e+01 1.42e-02 1.04e-01 6.00e-01
22 5.70e+01 1.31e+00 4.71e+03 4.31e+01 1.36e-02 1.02e-01 6.00e-01
23 5.52e+01 1.27e+00 4.56e+03 4.16e+01 1.31e-02 1.01e-01 6.00e-01
24 5.35e+01 1.24e+00 4.42e+03 4.03e+01 1.26e-02 9.98e-02 6.00e-01
25 5.18e+01 1.21e+00 4.28e+03 3.90e+01 1.22e-02 9.86e-02 6.00e-01
26 5.02e+01 1.18e+00 4.14e+03 3.78e+01 1.18e-02 9.75e-02 6.00e-01
27 4.87e+01 1.15e+00 4.02e+03 3.67e+01 1.14e-02 9.64e-02 6.00e-01
28 4.72e+01 1.12e+00 3.89e+03 3.56e+01 1.10e-02 9.55e-02 6.00e-01
29 4.57e+01 1.09e+00 3.77e+03 3.46e+01 1.07e-02 9.45e-02 6.00e-01
30 4.44e+01 1.06e+00 3.66e+03 3.36e+01 1.03e-02 9.36e-02 6.00e-01
31 4.31e+01 1.03e+00 3.56e+03 3.27e+01 1.00e-02 9.26e-02 6.00e-01
32 4.20e+01 1.01e+00 3.46e+03 3.18e+01 9.72e-03 9.16e-02 6.00e-01
33 4.09e+01 9.83e-01 3.37e+03 3.09e+01 9.44e-03 9.06e-02 6.00e-01
34 3.98e+01 9.58e-01 3.28e+03 3.02e+01 9.17e-03 8.97e-02 6.00e-01
35 3.88e+01 9.33e-01 3.20e+03 2.94e+01 8.91e-03 8.87e-02 6.00e-01
36 3.79e+01 9.08e-01 3.12e+03 2.87e+01 8.66e-03 8.77e-02 6.00e-01
37 3.70e+01 8.84e-01 3.05e+03 2.80e+01 8.43e-03 8.67e-02 6.00e-01
38 3.62e+01 8.60e-01 2.99e+03 2.74e+01 8.21e-03 8.57e-02 6.00e-01
39 3.54e+01 8.37e-01 2.92e+03 2.68e+01 7.99e-03 8.46e-02 6.00e-01
40 3.47e+01 8.14e-01 2.87e+03 2.62e+01 7.79e-03 8.35e-02 6.00e-01
41 3.41e+01 7.92e-01 2.81e+03 2.57e+01 7.59e-03 8.24e-02 6.00e-01
42 3.35e+01 7.71e-01 2.76e+03 2.52e+01 7.40e-03 8.13e-02 6.00e-01
43 3.29e+01 7.50e-01 2.72e+03 2.47e+01 7.22e-03 8.01e-02 6.00e-01
44 3.23e+01 7.31e-01 2.67e+03 2.42e+01 7.04e-03 7.90e-02 6.00e-01
45 3.17e+01 7.12e-01 2.63e+03 2.38e+01 6.87e-03 7.78e-02 6.00e-01
46 3.12e+01 6.94e-01 2.58e+03 2.34e+01 6.71e-03 7.66e-02 6.00e-01
47 3.07e+01 6.77e-01 2.54e+03 2.30e+01 6.54e-03 7.54e-02 6.00e-01
48 3.02e+01 6.61e-01 2.50e+03 2.27e+01 6.39e-03 7.42e-02 6.00e-01
49 2.98e+01 6.46e-01 2.46e+03 2.24e+01 6.23e-03 7.30e-02 6.00e-01
50 2.93e+01 6.32e-01 2.43e+03 2.21e+01 6.08e-03 7.18e-02 6.00e-01
51 2.89e+01 6.18e-01 2.39e+03 2.18e+01 5.94e-03 7.05e-02 6.00e-01
52 2.84e+01 6.06e-01 2.35e+03 2.15e+01 5.79e-03 6.93e-02 6.00e-01
53 2.80e+01 5.94e-01 2.31e+03 2.13e+01 5.65e-03 6.81e-02 6.00e-01
54 2.76e+01 5.83e-01 2.28e+03 2.11e+01 5.52e-03 6.69e-02 6.00e-01
55 2.72e+01 5.72e-01 2.24e+03 2.09e+01 5.39e-03 6.57e-02 6.00e-01
56 2.68e+01 5.63e-01 2.21e+03 2.07e+01 5.26e-03 6.45e-02 6.00e-01
57 2.64e+01 5.53e-01 2.17e+03 2.05e+01 5.13e-03 6.33e-02 6.00e-01
58 2.60e+01 5.44e-01 2.14e+03 2.04e+01 5.01e-03 6.20e-02 6.00e-01
59 2.57e+01 5.36e-01 2.11e+03 2.03e+01 4.89e-03 6.08e-02 6.00e-01
60 2.53e+01 5.28e-01 2.07e+03 2.01e+01 4.78e-03 5.97e-02 6.00e-01
61 2.50e+01 5.21e-01 2.04e+03 2.00e+01 4.66e-03 5.85e-02 6.00e-01
62 2.46e+01 5.13e-01 2.01e+03 1.99e+01 4.56e-03 5.73e-02 6.00e-01
63 2.43e+01 5.07e-01 1.98e+03 1.98e+01 4.45e-03 5.62e-02 6.00e-01
64 2.40e+01 5.00e-01 1.96e+03 1.97e+01 4.35e-03 5.51e-02 6.00e-01
65 2.37e+01 4.94e-01 1.93e+03 1.97e+01 4.25e-03 5.39e-02 6.00e-01
66 2.34e+01 4.88e-01 1.90e+03 1.96e+01 4.15e-03 5.29e-02 6.00e-01
67 2.32e+01 4.82e-01 1.88e+03 1.95e+01 4.06e-03 5.18e-02 6.00e-01
68 2.29e+01 4.76e-01 1.86e+03 1.95e+01 3.97e-03 5.07e-02 6.00e-01
69 2.27e+01 4.71e-01 1.83e+03 1.94e+01 3.88e-03 4.97e-02 6.00e-01
70 2.25e+01 4.66e-01 1.81e+03 1.94e+01 3.80e-03 4.86e-02 6.00e-01
71 2.22e+01 4.61e-01 1.79e+03 1.93e+01 3.72e-03 4.76e-02 6.00e-01
72 2.20e+01 4.56e-01 1.77e+03 1.93e+01 3.64e-03 4.67e-02 6.00e-01
73 2.18e+01 4.52e-01 1.75e+03 1.92e+01 3.56e-03 4.57e-02 6.00e-01
74 2.16e+01 4.48e-01 1.73e+03 1.92e+01 3.48e-03 4.48e-02 6.00e-01
75 2.14e+01 4.44e-01 1.71e+03 1.92e+01 3.41e-03 4.39e-02 6.00e-01
76 2.12e+01 4.40e-01 1.69e+03 1.91e+01 3.34e-03 4.30e-02 6.00e-01
77 2.10e+01 4.37e-01 1.67e+03 1.91e+01 3.27e-03 4.21e-02 6.00e-01
78 2.08e+01 4.34e-01 1.65e+03 1.91e+01 3.20e-03 4.13e-02 6.00e-01
79 2.06e+01 4.31e-01 1.63e+03 1.90e+01 3.13e-03 4.05e-02 6.00e-01
80 2.04e+01 4.28e-01 1.62e+03 1.90e+01 3.07e-03 3.97e-02 6.00e-01
81 2.02e+01 4.25e-01 1.60e+03 1.90e+01 3.00e-03 3.89e-02 6.00e-01
82 2.00e+01 4.22e-01 1.58e+03 1.89e+01 2.94e-03 3.81e-02 6.00e-01
83 1.98e+01 4.20e-01 1.56e+03 1.89e+01 2.88e-03 3.74e-02 6.00e-01
84 1.97e+01 4.18e-01 1.55e+03 1.89e+01 2.82e-03 3.67e-02 6.00e-01
85 1.95e+01 4.16e-01 1.53e+03 1.88e+01 2.76e-03 3.60e-02 6.00e-01
86 1.93e+01 4.14e-01 1.52e+03 1.88e+01 2.70e-03 3.54e-02 6.00e-01
87 1.92e+01 4.12e-01 1.50e+03 1.88e+01 2.65e-03 3.47e-02 6.00e-01
88 1.90e+01 4.10e-01 1.48e+03 1.87e+01 2.59e-03 3.41e-02 6.00e-01
89 1.88e+01 4.08e-01 1.47e+03 1.87e+01 2.54e-03 3.35e-02 6.00e-01
90 1.87e+01 4.07e-01 1.45e+03 1.87e+01 2.49e-03 3.30e-02 6.00e-01
91 1.85e+01 4.05e-01 1.44e+03 1.87e+01 2.44e-03 3.24e-02 6.00e-01
92 1.84e+01 4.04e-01 1.43e+03 1.86e+01 2.39e-03 3.18e-02 6.00e-01
93 1.82e+01 4.02e-01 1.41e+03 1.86e+01 2.34e-03 3.13e-02 6.00e-01
94 1.81e+01 4.01e-01 1.40e+03 1.86e+01 2.29e-03 3.07e-02 6.00e-01
95 1.80e+01 4.00e-01 1.39e+03 1.85e+01 2.25e-03 3.02e-02 6.00e-01
96 1.78e+01 3.98e-01 1.37e+03 1.85e+01 2.20e-03 2.96e-02 6.00e-01
97 1.77e+01 3.97e-01 1.36e+03 1.84e+01 2.16e-03 2.91e-02 6.00e-01
98 1.76e+01 3.96e-01 1.35e+03 1.84e+01 2.12e-03 2.85e-02 6.00e-01
99 1.75e+01 3.95e-01 1.34e+03 1.84e+01 2.08e-03 2.80e-02 6.00e-01
100 1.74e+01 3.94e-01 1.33e+03 1.83e+01 2.04e-03 2.74e-02 6.00e-01
101 1.73e+01 3.93e-01 1.32e+03 1.83e+01 2.00e-03 2.69e-02 6.00e-01
102 1.72e+01 3.92e-01 1.32e+03 1.83e+01 1.96e-03 2.63e-02 6.00e-01
103 1.71e+01 3.91e-01 1.31e+03 1.82e+01 1.93e-03 2.58e-02 6.00e-01
104 1.70e+01 3.90e-01 1.30e+03 1.82e+01 1.89e-03 2.52e-02 6.00e-01
105 1.70e+01 3.89e-01 1.29e+03 1.82e+01 1.86e-03 2.47e-02 6.00e-01
106 1.69e+01 3.88e-01 1.29e+03 1.81e+01 1.82e-03 2.42e-02 6.00e-01
107 1.68e+01 3.88e-01 1.28e+03 1.81e+01 1.79e-03 2.37e-02 6.00e-01
108 1.67e+01 3.87e-01 1.27e+03 1.80e+01 1.76e-03 2.32e-02 6.00e-01
109 1.66e+01 3.87e-01 1.27e+03 1.80e+01 1.73e-03 2.27e-02 6.00e-01
110 1.66e+01 3.86e-01 1.26e+03 1.80e+01 1.69e-03 2.23e-02 6.00e-01
111 1.65e+01 3.86e-01 1.25e+03 1.79e+01 1.66e-03 2.18e-02 6.00e-01
112 1.64e+01 3.85e-01 1.24e+03 1.79e+01 1.63e-03 2.14e-02 6.00e-01
113 1.63e+01 3.85e-01 1.24e+03 1.79e+01 1.60e-03 2.10e-02 6.00e-01
114 1.62e+01 3.85e-01 1.23e+03 1.78e+01 1.57e-03 2.07e-02 6.00e-01
115 1.62e+01 3.84e-01 1.22e+03 1.78e+01 1.54e-03 2.03e-02 6.00e-01
116 1.61e+01 3.84e-01 1.21e+03 1.78e+01 1.51e-03 2.00e-02 6.00e-01
117 1.60e+01 3.84e-01 1.21e+03 1.78e+01 1.48e-03 1.97e-02 6.00e-01
118 1.59e+01 3.84e-01 1.20e+03 1.77e+01 1.45e-03 1.94e-02 6.00e-01
119 1.58e+01 3.83e-01 1.19e+03 1.77e+01 1.43e-03 1.91e-02 6.00e-01
120 1.58e+01 3.83e-01 1.19e+03 1.77e+01 1.40e-03 1.88e-02 6.00e-01
121 1.57e+01 3.83e-01 1.18e+03 1.76e+01 1.37e-03 1.85e-02 6.00e-01
122 1.56e+01 3.83e-01 1.17e+03 1.76e+01 1.35e-03 1.82e-02 6.00e-01
123 1.56e+01 3.83e-01 1.17e+03 1.76e+01 1.32e-03 1.79e-02 6.00e-01
124 1.55e+01 3.83e-01 1.16e+03 1.76e+01 1.30e-03 1.77e-02 6.00e-01
125 1.54e+01 3.83e-01 1.16e+03 1.75e+01 1.28e-03 1.74e-02 6.00e-01
126 1.54e+01 3.82e-01 1.15e+03 1.75e+01 1.26e-03 1.71e-02 6.00e-01
127 1.53e+01 3.82e-01 1.15e+03 1.75e+01 1.23e-03 1.69e-02 6.00e-01
128 1.53e+01 3.82e-01 1.14e+03 1.75e+01 1.21e-03 1.66e-02 6.00e-01
129 1.52e+01 3.82e-01 1.14e+03 1.74e+01 1.19e-03 1.64e-02 6.00e-01
130 1.52e+01 3.82e-01 1.13e+03 1.74e+01 1.17e-03 1.61e-02 6.00e-01
131 1.51e+01 3.82e-01 1.13e+03 1.74e+01 1.16e-03 1.59e-02 6.00e-01
132 1.51e+01 3.81e-01 1.12e+03 1.74e+01 1.14e-03 1.56e-02 6.00e-01
133 1.50e+01 3.81e-01 1.12e+03 1.74e+01 1.12e-03 1.54e-02 6.00e-01
134 1.50e+01 3.81e-01 1.11e+03 1.73e+01 1.10e-03 1.51e-02 6.00e-01
135 1.49e+01 3.81e-01 1.11e+03 1.73e+01 1.09e-03 1.49e-02 6.00e-01
136 1.49e+01 3.81e-01 1.11e+03 1.73e+01 1.07e-03 1.47e-02 6.00e-01
137 1.49e+01 3.80e-01 1.10e+03 1.73e+01 1.05e-03 1.45e-02 6.00e-01
138 1.48e+01 3.80e-01 1.10e+03 1.72e+01 1.04e-03 1.43e-02 6.00e-01
139 1.48e+01 3.80e-01 1.09e+03 1.72e+01 1.02e-03 1.41e-02 6.00e-01
140 1.47e+01 3.80e-01 1.09e+03 1.72e+01 1.01e-03 1.39e-02 6.00e-01
141 1.47e+01 3.79e-01 1.09e+03 1.72e+01 9.91e-04 1.37e-02 6.00e-01
142 1.46e+01 3.79e-01 1.08e+03 1.72e+01 9.76e-04 1.35e-02 6.00e-01
143 1.46e+01 3.79e-01 1.08e+03 1.71e+01 9.62e-04 1.33e-02 6.00e-01
144 1.46e+01 3.79e-01 1.08e+03 1.71e+01 9.48e-04 1.31e-02 6.00e-01
145 1.45e+01 3.79e-01 1.07e+03 1.71e+01 9.34e-04 1.30e-02 6.00e-01
146 1.45e+01 3.78e-01 1.07e+03 1.71e+01 9.20e-04 1.28e-02 6.00e-01
147 1.44e+01 3.78e-01 1.07e+03 1.71e+01 9.06e-04 1.26e-02 6.00e-01
148 1.44e+01 3.78e-01 1.06e+03 1.70e+01 8.93e-04 1.25e-02 6.00e-01
149 1.44e+01 3.78e-01 1.06e+03 1.70e+01 8.79e-04 1.23e-02 6.00e-01
150 1.43e+01 3.77e-01 1.06e+03 1.70e+01 8.66e-04 1.21e-02 6.00e-01
151 1.43e+01 3.77e-01 1.05e+03 1.70e+01 8.54e-04 1.20e-02 6.00e-01
152 1.43e+01 3.77e-01 1.05e+03 1.69e+01 8.41e-04 1.18e-02 6.00e-01
153 1.42e+01 3.77e-01 1.05e+03 1.69e+01 8.29e-04 1.17e-02 6.00e-01
154 1.42e+01 3.77e-01 1.04e+03 1.69e+01 8.16e-04 1.16e-02 6.00e-01
155 1.42e+01 3.77e-01 1.04e+03 1.69e+01 8.05e-04 1.14e-02 6.00e-01
156 1.41e+01 3.76e-01 1.04e+03 1.69e+01 7.93e-04 1.13e-02 6.00e-01
157 1.41e+01 3.76e-01 1.03e+03 1.68e+01 7.82e-04 1.12e-02 6.00e-01
158 1.41e+01 3.76e-01 1.03e+03 1.68e+01 7.70e-04 1.10e-02 6.00e-01
159 1.40e+01 3.76e-01 1.03e+03 1.68e+01 7.59e-04 1.09e-02 6.00e-01
160 1.40e+01 3.76e-01 1.03e+03 1.68e+01 7.49e-04 1.08e-02 6.00e-01
161 1.40e+01 3.76e-01 1.02e+03 1.67e+01 7.38e-04 1.07e-02 6.00e-01
162 1.39e+01 3.76e-01 1.02e+03 1.67e+01 7.28e-04 1.05e-02 6.00e-01
163 1.39e+01 3.76e-01 1.02e+03 1.67e+01 7.18e-04 1.04e-02 6.00e-01
164 1.39e+01 3.75e-01 1.02e+03 1.67e+01 7.09e-04 1.03e-02 6.00e-01
165 1.39e+01 3.75e-01 1.01e+03 1.67e+01 6.99e-04 1.02e-02 6.00e-01
166 1.38e+01 3.75e-01 1.01e+03 1.66e+01 6.90e-04 1.01e-02 6.00e-01
167 1.38e+01 3.75e-01 1.01e+03 1.66e+01 6.81e-04 9.96e-03 6.00e-01
168 1.38e+01 3.75e-01 1.01e+03 1.66e+01 6.73e-04 9.85e-03 6.00e-01
169 1.38e+01 3.75e-01 1.01e+03 1.66e+01 6.64e-04 9.74e-03 6.00e-01
170 1.37e+01 3.75e-01 1.00e+03 1.66e+01 6.56e-04 9.63e-03 6.00e-01
171 1.37e+01 3.75e-01 1.00e+03 1.65e+01 6.48e-04 9.52e-03 6.00e-01
172 1.37e+01 3.75e-01 1.00e+03 1.65e+01 6.39e-04 9.41e-03 6.00e-01
173 1.37e+01 3.75e-01 1.00e+03 1.65e+01 6.32e-04 9.30e-03 6.00e-01
174 1.36e+01 3.75e-01 9.98e+02 1.65e+01 6.24e-04 9.19e-03 6.00e-01
175 1.36e+01 3.75e-01 9.96e+02 1.65e+01 6.16e-04 9.08e-03 6.00e-01
176 1.36e+01 3.75e-01 9.95e+02 1.64e+01 6.08e-04 8.98e-03 6.00e-01
177 1.36e+01 3.74e-01 9.93e+02 1.64e+01 6.01e-04 8.88e-03 6.00e-01
178 1.36e+01 3.74e-01 9.92e+02 1.64e+01 5.93e-04 8.78e-03 6.00e-01
179 1.36e+01 3.74e-01 9.90e+02 1.64e+01 5.86e-04 8.68e-03 6.00e-01
180 1.35e+01 3.74e-01 9.88e+02 1.64e+01 5.78e-04 8.59e-03 6.00e-01
181 1.35e+01 3.74e-01 9.87e+02 1.64e+01 5.71e-04 8.49e-03 6.00e-01
182 1.35e+01 3.74e-01 9.85e+02 1.63e+01 5.64e-04 8.40e-03 6.00e-01
183 1.35e+01 3.74e-01 9.84e+02 1.63e+01 5.56e-04 8.31e-03 6.00e-01
184 1.35e+01 3.74e-01 9.82e+02 1.63e+01 5.49e-04 8.23e-03 6.00e-01
185 1.34e+01 3.74e-01 9.80e+02 1.63e+01 5.42e-04 8.14e-03 6.00e-01
186 1.34e+01 3.74e-01 9.79e+02 1.63e+01 5.35e-04 8.06e-03 6.00e-01
187 1.34e+01 3.74e-01 9.77e+02 1.63e+01 5.28e-04 7.98e-03 6.00e-01
188 1.34e+01 3.74e-01 9.76e+02 1.62e+01 5.22e-04 7.90e-03 6.00e-01
189 1.34e+01 3.74e-01 9.74e+02 1.62e+01 5.15e-04 7.83e-03 6.00e-01
190 1.33e+01 3.74e-01 9.73e+02 1.62e+01 5.09e-04 7.75e-03 6.00e-01
191 1.33e+01 3.74e-01 9.71e+02 1.62e+01 5.02e-04 7.67e-03 6.00e-01
192 1.33e+01 3.74e-01 9.70e+02 1.62e+01 4.96e-04 7.59e-03 6.00e-01
193 1.33e+01 3.73e-01 9.69e+02 1.62e+01 4.90e-04 7.52e-03 6.00e-01
194 1.33e+01 3.73e-01 9.67e+02 1.62e+01 4.84e-04 7.44e-03 6.00e-01
195 1.33e+01 3.73e-01 9.66e+02 1.61e+01 4.79e-04 7.37e-03 6.00e-01
196 1.32e+01 3.73e-01 9.65e+02 1.61e+01 4.73e-04 7.29e-03 6.00e-01
197 1.32e+01 3.73e-01 9.64e+02 1.61e+01 4.67e-04 7.21e-03 6.00e-01
198 1.32e+01 3.73e-01 9.63e+02 1.61e+01 4.62e-04 7.14e-03 6.00e-01
199 1.32e+01 3.73e-01 9.62e+02 1.61e+01 4.57e-04 7.06e-03 6.00e-01
--------------------------------------------------------------------------
Display solve time and reconstruction performance.
print("AddMaskSim wrapped ConvBPDN solve time: %.2fs" % t)
print("Corrupted image PSNR: %5.2f dB" % metric.psnr(img, imgw))
print("Recovered image PSNR: %5.2f dB" % metric.psnr(img, imgr))
AddMaskSim wrapped ConvBPDN solve time: 1.57s
Corrupted image PSNR: 9.10 dB
Recovered image PSNR: 25.51 dB
Display reference, test, and reconstructed image
fig = plot.figure(figsize=(21, 7))
plot.subplot(1, 3, 1)
plot.imview(img, title='Reference image', fig=fig)
plot.subplot(1, 3, 2)
plot.imview(imgw, title='Corrupted image', fig=fig)
plot.subplot(1, 3, 3)
plot.imview(imgr, title='Reconstructed image', fig=fig)
fig.show()
Display lowpass component and sparse representation
fig = plot.figure(figsize=(14, 7))
plot.subplot(1, 2, 1)
plot.imview(X[..., 0], cmap=plot.cm.Blues, title='Lowpass component', fig=fig)
plot.subplot(1, 2, 2)
plot.imview(np.sum(abs(X[..., 1:]).squeeze(), axis=-1), cmap=plot.cm.Blues,
title='Sparse representation', fig=fig)
fig.show()
Plot functional value, residuals, and rho (not available if GPU implementation used).
if ams is not None:
its = ams.getitstat()
fig = plot.figure(figsize=(21, 7))
plot.subplot(1, 3, 1)
plot.plot(its.ObjFun, xlbl='Iterations', ylbl='Functional', fig=fig)
plot.subplot(1, 3, 2)
plot.plot(np.vstack((its.PrimalRsdl, its.DualRsdl)).T, ptyp='semilogy',
xlbl='Iterations', ylbl='Residual', lgnd=['Primal', 'Dual'],
fig=fig)
plot.subplot(1, 3, 3)
plot.plot(its.Rho, xlbl='Iterations', ylbl='Penalty Parameter', fig=fig)
fig.show()