def test_ssim(space):
ground_truth = odl.phantom.white_noise(space)
# SSIM of true image should be either
# * 1 with force_lower_is_better == False,
# * -1 with force_lower_is_better == True and normalized == False,
# * 0 with force_lower_is_better == True and normalized == True.
result = fom.ssim(ground_truth, ground_truth)
assert result == pytest.approx(1)
result_normalized = fom.ssim(ground_truth, ground_truth, normalized=True)
assert result_normalized == pytest.approx(1)
result_flib = fom.ssim(ground_truth,
ground_truth,
force_lower_is_better=True)
assert result_flib == pytest.approx(-1)
result_nf = fom.ssim(ground_truth,
ground_truth,
normalized=True,
force_lower_is_better=True)
assert result_nf == pytest.approx(0)
# SSIM with ground truth zero should always give zero if not normalized
# and 1/2 otherwise.
data = odl.phantom.white_noise(space)
result = fom.ssim(data, space.zero())
assert result == pytest.approx(0)
result_normalized = fom.ssim(data, space.zero(), normalized=True)
assert result_normalized == pytest.approx(0.5)
result_flib = fom.ssim(data, space.zero(), force_lower_is_better=True)
assert result_flib == pytest.approx(0)
result_nf = fom.ssim(data,
space.zero(),
normalized=True,
force_lower_is_better=True)
assert result_nf == pytest.approx(0.5)
# SSIM should be symmetric if the dynamic range is set explicitly.
for nor in [True, False]:
for flib in [True, False]:
result1 = fom.ssim(data,
ground_truth,
dynamic_range=1,
normalized=nor,
force_lower_is_better=flib)
result2 = fom.ssim(ground_truth,
data,
dynamic_range=1,
normalized=nor,
force_lower_is_better=flib)
assert result1 == pytest.approx(result2)
def __call__(self, x, **kwargs):
if len(x) == 2:
x = x[0]
k = self.iter_count
if k in self.iter_save:
if self.obj_fun is not None:
self.obj.append(self.obj_fun(x))
if k in self.iter_plot:
name = '{}{:04d}'.format(self.prefix, k)
gtruth = self.gtruth
if gtruth is not None:
name += '_psnr{:.1f}db_ssim{:0.4f}'.format(
psnr(x, gtruth), ssim(x, gtruth))
save_result(x, name)
if self.error and gtruth is not None:
save_error(x - gtruth, name + '_error')
self.iter_count += 1
# saveto=(save_path + 'cp_iter_{}').format(imax_val))
fig = plt.figure()
plt.imshow(np.rot90(x_rec_odl.asarray()), cmap='bone', clim=[0, 2.33])
plt.xticks([])
plt.yticks([])
fig.savefig((save_path + 'cp_iter_{}').format(imax_val),
transparent=True,
bbox_inches='tight',
pad_inches=0)
fig = plt.figure()
plt.imshow(np.rot90(x_rec_odl.asarray()), cmap='bone', clim=[0.8, 1.2])
plt.xticks([])
plt.yticks([])
fig.savefig((save_path + 'cp_iter_{}_windowed').format(imax_val),
transparent=True,
bbox_inches='tight',
pad_inches=0)
mse_val = mean_squared_error(x_rec_odl, x_true_odl)
ssim_val = ssim(x_rec_odl, x_true_odl)
print('validation loss={}, data dis={}, regularizer={}'.format(
loss_result, data_disc_res, reg_res))
print('mse={}, ssim={}'.format(mse_val, ssim_val))
sys.stdout.log.close()
sys.stdout = sys.stdout.terminal
odl.solvers.douglas_rachford_pd(x,
f,
g,
L,
tau,
sigmas,
niter=200,
callback=callback)
# %% Show results
x_true.show('True image', force_show=True)
y.show('Noisy image', force_show=True)
x.show('Denoised image', force_show=True)
(x_true - x).show('Difference true - denoised', force_show=True)
# %% Compute some image quality metrics
print('Noisy')
print('-----')
print('Mean squared error:', fom.mean_squared_error(y, x_true))
print('PSNR:', fom.psnr(y, x_true))
print('SSIM:', fom.ssim(y, x_true))
print('')
print('Denoised')
print('--------')
print('Mean squared error:', fom.mean_squared_error(x, x_true))
print('PSNR:', fom.psnr(x, x_true))
print('SSIM:', fom.ssim(x, x_true))
blur.append(
fom.blurring(phantom_noisy,
phantom,
mask,
normalized=True,
smoothness_factor=30))
false_struct.append(
fom.false_structures(phantom_noisy,
phantom,
mask,
normalized=True,
smoothness_factor=30))
ssim.append(fom.ssim(phantom_noisy, phantom))
psnr.append(fom.psnr(phantom_noisy, phantom, normalized=True))
haarpsi.append(fom.haarpsi(phantom_noisy, phantom))
fig, ax = plt.subplots()
ax.plot(mse, label='MSE')
ax.plot(mae, label='MAE')
ax.plot(mvd, label='MVD')
ax.plot(std_diff, label='SDD')
ax.plot(range_diff, label='RD')
ax.plot(blur, label='BLUR')
ax.plot(false_struct, label='FS')
ax.plot(ssim, label='SSIM')
ax.plot(haarpsi, label='HaarPSI')
blur.append(
fom.blurring(phantom_noisy,
phantom,
mask,
normalized=True,
smoothness_factor=30))
false_struct.append(
fom.false_structures(phantom_noisy,
phantom,
mask,
normalized=True,
smoothness_factor=30))
ssim.append(fom.ssim(phantom_noisy, phantom, normalized=True))
plt.figure('Figures of merit')
plt.plot(mse, label='Mean squared error')
plt.plot(mae, label='Mean absolute error')
plt.plot(mvd, label='Mean value difference')
plt.plot(std_diff, label='Standard deviation difference')
plt.plot(range_diff, label='Range difference')
plt.plot(blur, label='Blurring')
plt.plot(false_struct, label='False structures')
plt.plot(ssim, label='Structural similarity')
plt.legend(loc='upper center',
bbox_to_anchor=(0.5, -0.15),
fancybox=True,
shadow=True,
