def test_2d_bf():
lx = 70
ly = 100
data, labels = make_2d_syntheticdata(lx, ly)
labels_bf = random_walker(data, labels, beta=90, mode="bf")
assert (labels_bf[25:45, 40:60] == 2).all()
full_prob_bf = random_walker(data, labels, beta=90, mode="bf", return_full_prob=True)
assert (full_prob_bf[1, 25:45, 40:60] >= full_prob_bf[0, 25:45, 40:60]).all()
return data, labels_bf, full_prob_bf
def test_length2_spacing():
# If this passes without raising an exception (warnings OK), the new
# spacing code is working properly.
np.random.seed(42)
img = np.ones((10, 10)) + 0.2 * np.random.normal(size=(10, 10))
labels = np.zeros((10, 10), dtype=np.uint8)
labels[2, 4] = 1
labels[6, 8] = 4
random_walker(img, labels, spacing=(1., 2.))
def test_2d_cg_mg():
lx = 70
ly = 100
data, labels = make_2d_syntheticdata(lx, ly)
labels_cg_mg = random_walker(data, labels, beta=90, mode="cg_mg")
assert (labels_cg_mg[25:45, 40:60] == 2).all()
full_prob = random_walker(data, labels, beta=90, mode="cg_mg", return_full_prob=True)
assert (full_prob[1, 25:45, 40:60] >= full_prob[0, 25:45, 40:60]).all()
return data, labels_cg_mg
def test_multispectral_2d():
lx, ly = 70, 100
data, labels = make_2d_syntheticdata(lx, ly)
data2 = data.copy()
data.shape += (1,)
data = data.repeat(2, axis=2) # Result should be identical
multi_labels = random_walker(data, labels, mode='cg', multichannel=True)
single_labels = random_walker(data2, labels, mode='cg')
assert (multi_labels.reshape(labels.shape)[25:45, 40:60] == 2).all()
return data, multi_labels, single_labels, labels
def test_multispectral_2d():
lx, ly = 70, 100
data, labels = make_2d_syntheticdata(lx, ly)
data = data[..., np.newaxis].repeat(2, axis=-1) # Expect identical output
multi_labels = random_walker(data, labels, mode='cg', multichannel=True)
assert data[..., 0].shape == labels.shape
single_labels = random_walker(data[..., 0], labels, mode='cg')
assert (multi_labels.reshape(labels.shape)[25:45, 40:60] == 2).all()
assert data[..., 0].shape == labels.shape
return data, multi_labels, single_labels, labels
def test_multispectral_3d():
n = 30
lx, ly, lz = n, n, n
data, labels = make_3d_syntheticdata(lx, ly, lz)
data.shape += (1,)
data = data.repeat(2, axis=3) # Result should be identical
multi_labels = random_walker(data, labels, mode='cg', multichannel=True)
single_labels = random_walker(data[..., 0], labels, mode='cg')
assert (multi_labels.reshape(labels.shape)[13:17, 13:17, 13:17] == 2).all()
assert (single_labels.reshape(labels.shape)[13:17, 13:17, 13:17] == 2).all()
return data, multi_labels, single_labels, labels
def test_multispectral_3d():
n = 30
lx, ly, lz = n, n, n
data, labels = make_3d_syntheticdata(lx, ly, lz)
data = data[..., np.newaxis].repeat(2, axis=-1) # Expect identical output
multi_labels = random_walker(data, labels, mode='cg', multichannel=True)
assert data[..., 0].shape == labels.shape
single_labels = random_walker(data[..., 0], labels, mode='cg')
assert (multi_labels.reshape(labels.shape)[13:17, 13:17, 13:17] == 2).all()
assert (single_labels.reshape(labels.shape)[13:17, 13:17, 13:17] == 2).all()
assert data[..., 0].shape == labels.shape
return data, multi_labels, single_labels, labels
def test_spacing_1():
n = 30
lx, ly, lz = n, n, n
data, _ = make_3d_syntheticdata(lx, ly, lz)
# Rescale `data` along Y axis
# `resize` is not yet 3D capable, so this must be done by looping in 2D.
data_aniso = np.zeros((n, n * 2, n))
for i, yz in enumerate(data):
data_aniso[i, :, :] = resize(yz, (n * 2, n),
mode='constant',
anti_aliasing=False)
# Generate new labels
small_l = int(lx // 5)
labels_aniso = np.zeros_like(data_aniso)
labels_aniso[lx // 5, ly // 5, lz // 5] = 1
labels_aniso[lx // 2 + small_l // 4,
ly - small_l // 2,
lz // 2 - small_l // 4] = 2
# Test with `spacing` kwarg
# First, anisotropic along Y
with expected_warnings(['"cg" mode' + '|' + SCIPY_RANK_WARNING,
NUMPY_MATRIX_WARNING]):
labels_aniso = random_walker(data_aniso, labels_aniso, mode='cg',
spacing=(1., 2., 1.))
assert (labels_aniso[13:17, 26:34, 13:17] == 2).all()
# Rescale `data` along X axis
# `resize` is not yet 3D capable, so this must be done by looping in 2D.
data_aniso = np.zeros((n, n * 2, n))
for i in range(data.shape[1]):
data_aniso[i, :, :] = resize(data[:, 1, :], (n * 2, n),
mode='constant',
anti_aliasing=False)
# Generate new labels
small_l = int(lx // 5)
labels_aniso2 = np.zeros_like(data_aniso)
labels_aniso2[lx // 5, ly // 5, lz // 5] = 1
labels_aniso2[lx - small_l // 2,
ly // 2 + small_l // 4,
lz // 2 - small_l // 4] = 2
# Anisotropic along X
with expected_warnings(['"cg" mode' + '|' + SCIPY_RANK_WARNING,
NUMPY_MATRIX_WARNING]):
labels_aniso2 = random_walker(data_aniso,
labels_aniso2,
mode='cg', spacing=(2., 1., 1.))
assert (labels_aniso2[26:34, 13:17, 13:17] == 2).all()
def test_2d_cg():
lx = 70
ly = 100
data, labels = make_2d_syntheticdata(lx, ly)
labels_cg = random_walker(data, labels, beta=90, mode='cg')
assert (labels_cg[25:45, 40:60] == 2).all()
assert data.shape == labels.shape
full_prob = random_walker(data, labels, beta=90, mode='cg',
return_full_prob=True)
assert (full_prob[1, 25:45, 40:60] >=
full_prob[0, 25:45, 40:60]).all()
assert data.shape == labels.shape
return data, labels_cg
def test_trivial_cases():
# When all voxels are labeled
img = np.ones((10, 10))
labels = np.ones((10, 10))
pass_through = random_walker(img, labels)
np.testing.assert_array_equal(pass_through, labels)
# When all voxels are labeled AND return_full_prob is True
labels[:, :5] = 3
expected = np.concatenate(((labels == 1)[..., np.newaxis],
(labels == 3)[..., np.newaxis]), axis=2)
test = random_walker(img, labels, return_full_prob=True)
np.testing.assert_array_equal(test, expected)
def test_2d_cg():
lx = 70
ly = 100
data, labels = make_2d_syntheticdata(lx, ly)
with expected_warnings(['"cg" mode' + '|' + SCIPY_EXPECTED]):
labels_cg = random_walker(data, labels, beta=90, mode='cg')
assert (labels_cg[25:45, 40:60] == 2).all()
assert data.shape == labels.shape
with expected_warnings(['"cg" mode' + '|' + SCIPY_EXPECTED]):
full_prob = random_walker(data, labels, beta=90, mode='cg',
return_full_prob=True)
assert (full_prob[1, 25:45, 40:60] >=
full_prob[0, 25:45, 40:60]).all()
assert data.shape == labels.shape
return data, labels_cg
def test_multispectral_2d():
lx, ly = 70, 100
data, labels = make_2d_syntheticdata(lx, ly)
data = data[..., np.newaxis].repeat(2, axis=-1) # Expect identical output
with expected_warnings(['"cg" mode' + '|' + SCIPY_RANK_WARNING,
NUMPY_MATRIX_WARNING]):
multi_labels = random_walker(data, labels, mode='cg',
multichannel=True)
assert data[..., 0].shape == labels.shape
with expected_warnings(['"cg" mode' + '|' + SCIPY_RANK_WARNING,
NUMPY_MATRIX_WARNING]):
single_labels = random_walker(data[..., 0], labels, mode='cg')
assert (multi_labels.reshape(labels.shape)[25:45, 40:60] == 2).all()
assert data[..., 0].shape == labels.shape
return data, multi_labels, single_labels, labels
def cluster_by_diffusion(data):
markers = np.zeros(data.shape, dtype=np.uint)
markers[data < -0.00] = 1
markers[data > 0.03] = 2
labels2 = random_walker(data, markers, beta=10, mode='bf')
return labels2
def start():
# Generate noisy synthetic data
data = microstructure(l=128)
data += 0.35 * np.random.randn(*data.shape)
markers = np.zeros(data.shape, dtype=np.uint)
markers[data < -0.3] = 1
markers[data > 1.3] = 2
# Run random walker algorithm
labels = random_walker(data, markers, beta=10, mode='bf')
# Plot results
fig, (ax1, ax2, ax3) = plt.subplots(1, 3, figsize=(8, 3.2))
ax1.imshow(data, cmap='gray', interpolation='nearest')
ax1.axis('off')
ax1.set_title('Noisy data')
ax2.imshow(markers, cmap='hot', interpolation='nearest')
ax2.axis('off')
ax2.set_title('Markers')
ax3.imshow(labels, cmap='gray', interpolation='nearest')
ax3.axis('off')
ax3.set_title('Segmentation')
fig.subplots_adjust(hspace=0.01, wspace=0.01, top=1, bottom=0, left=0,
right=1)
plt.show()
def test_2d_bf():
lx = 70
ly = 100
data, labels = make_2d_syntheticdata(lx, ly)
labels_bf = random_walker(data, labels, beta=90, mode="bf")
assert (labels_bf[25:45, 40:60] == 2).all()
assert data.shape == labels.shape
full_prob_bf = random_walker(data, labels, beta=90, mode="bf", return_full_prob=True)
assert (full_prob_bf[1, 25:45, 40:60] >= full_prob_bf[0, 25:45, 40:60]).all()
assert data.shape == labels.shape
# Now test with more than two labels
labels[55, 80] = 3
full_prob_bf = random_walker(data, labels, beta=90, mode="bf", return_full_prob=True)
assert (full_prob_bf[1, 25:45, 40:60] >= full_prob_bf[0, 25:45, 40:60]).all()
assert len(full_prob_bf) == 3
assert data.shape == labels.shape
def test_2d_cg_mg():
lx = 70
ly = 100
data, labels = make_2d_syntheticdata(lx, ly)
labels_cg_mg = random_walker(data, labels, beta=90, mode='cg_mg')
assert (labels_cg_mg[25:45, 40:60] == 2).all()
return data, labels_cg_mg
def segment_out_cells(base):
# TODO: try using OTSU for GFP thresholding
sel_elem = disk(2)
gfp_collector = np.sum(base, axis=0)
gfp_clustering_markers = np.zeros(gfp_collector.shape, dtype=np.uint8)
# random walker segment
gfp_clustering_markers[gfp_collector > np.mean(gfp_collector) * 2] = 2
gfp_clustering_markers[gfp_collector < np.mean(gfp_collector) * 0.20] = 1
labels = random_walker(gfp_collector, gfp_clustering_markers, beta=10, mode='bf')
# round up the labels and set the background to 0 from 1
labels = closing(labels, sel_elem)
labels -= 1
# prepare distances for the watershed
distance = ndi.distance_transform_edt(labels)
local_maxi = peak_local_max(distance,
indices=False, # we want the image mask, not peak position
min_distance=10, # about half of a bud with our size
threshold_abs=10, # allows to clear the noise
labels=labels)
# we fuse the labels that are close together that escaped the min distance in local_maxi
local_maxi = ndi.convolve(local_maxi, np.ones((5, 5)), mode='constant', cval=0.0)
# finish the watershed
expanded_maxi_markers = ndi.label(local_maxi, structure=np.ones((3, 3)))[0]
segmented_cells_labels = watershed(-distance, expanded_maxi_markers, mask=labels)
# log debugging data
running_debug_frame.gfp_collector = gfp_collector
running_debug_frame.gfp_clustering_markers = gfp_clustering_markers
running_debug_frame.labels = labels
running_debug_frame.segmented_cells_labels = segmented_cells_labels
return gfp_collector, segmented_cells_labels
def _automatic_localization(self):
"""
Automatic localization made by Tomas Ryba.
"""
# seeds = self.get_seeds_using_class_1(class1)
liver = self.data3d * (self.segmentation != 0)
print('analyzing histogram...')
class1 = tools.analyse_histogram(self.data3d,
roi=self.segmentation != 0)
# sed3.sed3(self.data3d, seeds=class1).show()
print('getting seeds...')
seeds = self.get_seeds_using_prob_class1(
liver,
class1,
thresholdType='percOfMaxDist',
percT=0.3)
# sed3.sed3(self.data3d, seeds=seeds).show()
print('Starting random walker...')
rw = random_walker(liver, seeds, mode='cg_mg')
print('...finished.')
label_l = self.data['slab']['lesions']
lessions = rw == 2
sed3.sed3(self.data3d, contour=lessions).show()
lessions = self.filter_objects(lessions)
self.segmentation = np.where(lessions, label_l, self.segmentation)
def test_3d():
n = 30
lx, ly, lz = n, n, n
data, labels = make_3d_syntheticdata(lx, ly, lz)
labels = random_walker(data, labels, mode='cg')
assert (labels.reshape(data.shape)[13:17, 13:17, 13:17] == 2).all()
return data, labels
def test_2d_cg_mg():
lx = 70
ly = 100
data, labels = make_2d_syntheticdata(lx, ly)
expected = 'scipy.sparse.sparsetools|%s' % PYAMG_SCIPY_EXPECTED
with expected_warnings([expected]):
labels_cg_mg = random_walker(data, labels, beta=90, mode='cg_mg')
assert (labels_cg_mg[25:45, 40:60] == 2).all()
assert data.shape == labels.shape
with expected_warnings([expected]):
full_prob = random_walker(data, labels, beta=90, mode='cg_mg',
return_full_prob=True)
assert (full_prob[1, 25:45, 40:60] >=
full_prob[0, 25:45, 40:60]).all()
assert data.shape == labels.shape
return data, labels_cg_mg
def test_spacing_0():
n = 30
lx, ly, lz = n, n, n
data, _ = make_3d_syntheticdata(lx, ly, lz)
# Rescale `data` along Z axis
data_aniso = np.zeros((n, n, n // 2))
for i, yz in enumerate(data):
data_aniso[i, :, :] = resize(yz, (n, n // 2),
mode='constant',
anti_aliasing=False)
# Generate new labels
small_l = int(lx // 5)
labels_aniso = np.zeros_like(data_aniso)
labels_aniso[lx // 5, ly // 5, lz // 5] = 1
labels_aniso[lx // 2 + small_l // 4,
ly // 2 - small_l // 4,
lz // 4 - small_l // 8] = 2
# Test with `spacing` kwarg
with expected_warnings(['"cg" mode' + '|' + SCIPY_RANK_WARNING,
NUMPY_MATRIX_WARNING]):
labels_aniso = random_walker(data_aniso, labels_aniso, mode='cg',
spacing=(1., 1., 0.5))
assert (labels_aniso[13:17, 13:17, 7:9] == 2).all()
def test_reorder_labels():
lx = 70
ly = 100
data, labels = make_2d_syntheticdata(lx, ly)
labels[labels == 2] = 4
labels_bf = random_walker(data, labels, beta=90, mode='bf')
assert (labels_bf[25:45, 40:60] == 2).all()
return data, labels_bf
def test_types():
lx = 70
ly = 100
data, labels = make_2d_syntheticdata(lx, ly)
data = 255 * (data - data.min()) / (data.max() - data.min())
data = data.astype(np.uint8)
labels_cg_mg = random_walker(data, labels, beta=90, mode='cg_mg')
assert (labels_cg_mg[25:45, 40:60] == 2).all()
return data, labels_cg_mg
def test_2d_inactive():
lx = 70
ly = 100
data, labels = make_2d_syntheticdata(lx, ly)
labels[10:20, 10:20] = -1
labels[46:50, 33:38] = -2
labels = random_walker(data, labels, beta=90)
assert (labels.reshape((lx, ly))[25:45, 40:60] == 2).all()
return data, labels
def test_3d():
n = 30
lx, ly, lz = n, n, n
data, labels = make_3d_syntheticdata(lx, ly, lz)
with expected_warnings(['"cg" mode' + '|' + SCIPY_EXPECTED]):
labels = random_walker(data, labels, mode='cg')
assert (labels.reshape(data.shape)[13:17, 13:17, 13:17] == 2).all()
assert data.shape == labels.shape
return data, labels
def test_isolated_seeds():
np.random.seed(0)
a = np.random.random((7, 7))
mask = - np.ones(a.shape)
# This pixel is an isolated seed
mask[1, 1] = 1
# Unlabeled pixels
mask[3:, 3:] = 0
# Seeds connected to unlabeled pixels
mask[4, 4] = 2
mask[6, 6] = 1
# Test that no error is raised, and that labels of isolated seeds are OK
res = random_walker(a, mask)
assert res[1, 1] == 1
res = random_walker(a, mask, return_full_prob=True)
assert res[0, 1, 1] == 1
assert res[1, 1, 1] == 0
def test_2d_cg_mg():
lx = 70
ly = 100
data, labels = make_2d_syntheticdata(lx, ly)
anticipated_warnings = [
'scipy.sparse.sparsetools|%s' % PYAMG_OR_SCIPY_WARNING,
NUMPY_MATRIX_WARNING]
with expected_warnings(anticipated_warnings):
labels_cg_mg = random_walker(data, labels, beta=90, mode='cg_mg')
assert (labels_cg_mg[25:45, 40:60] == 2).all()
assert data.shape == labels.shape
with expected_warnings(anticipated_warnings):
full_prob = random_walker(data, labels, beta=90, mode='cg_mg',
return_full_prob=True)
assert (full_prob[1, 25:45, 40:60] >=
full_prob[0, 25:45, 40:60]).all()
assert data.shape == labels.shape
return data, labels_cg_mg
def test_3d_inactive():
n = 30
lx, ly, lz = n, n, n
data, labels = make_3d_syntheticdata(lx, ly, lz)
old_labels = np.copy(labels)
labels[5:25, 26:29, 26:29] = -1
after_labels = np.copy(labels)
labels = random_walker(data, labels, mode='cg')
assert (labels.reshape(data.shape)[13:17, 13:17, 13:17] == 2).all()
return data, labels, old_labels, after_labels
def test_reorder_labels():
lx = 70
ly = 100
data, labels = make_2d_syntheticdata(lx, ly)
labels[labels == 2] = 4
with expected_warnings([NUMPY_MATRIX_WARNING]):
labels_bf = random_walker(data, labels, beta=90, mode='bf')
assert (labels_bf[25:45, 40:60] == 2).all()
assert data.shape == labels.shape
return data, labels_bf
def test_types():
lx = 70
ly = 100
data, labels = make_2d_syntheticdata(lx, ly)
data = 255 * (data - data.min()) // (data.max() - data.min())
data = data.astype(np.uint8)
with expected_warnings([PYAMG_SCIPY_EXPECTED]):
labels_cg_mg = random_walker(data, labels, beta=90, mode='cg_mg')
assert (labels_cg_mg[25:45, 40:60] == 2).all()
assert data.shape == labels.shape
return data, labels_cg_mg
def binarize_img(im, thrs1, thrs2, max_iter):
start_time = time.time()
markers = np.zeros_like(im)
markers[im > thrs1] = 1
markers[im < thrs2] = 2
t = random_walker(im, markers, beta=100)
global count
print("count", count, " of ", max_iter, f" %: {count/max_iter:.2f}",
"| time (min): ", (time.time() - start_time) / 60)
count += 1
return t < 2
def test_multispectral_3d(dtype):
n = 30
lx, ly, lz = n, n, n
data, labels = make_3d_syntheticdata(lx, ly, lz)
data = data.astype(dtype, copy=False)
data = data[..., np.newaxis].repeat(2, axis=-1) # Expect identical output
with expected_warnings(
['"cg" mode' + '|' + SCIPY_RANK_WARNING, NUMPY_MATRIX_WARNING]):
multi_labels = random_walker(data,
labels,
mode='cg',
multichannel=True)
assert data[..., 0].shape == labels.shape
with expected_warnings(
['"cg" mode' + '|' + SCIPY_RANK_WARNING, NUMPY_MATRIX_WARNING]):
single_labels = random_walker(data[..., 0], labels, mode='cg')
assert (multi_labels.reshape(labels.shape)[13:17, 13:17, 13:17] == 2).all()
assert (single_labels.reshape(labels.shape)[13:17, 13:17,
13:17] == 2).all()
assert data[..., 0].shape == labels.shape
return data, multi_labels, single_labels, labels
def test_2d_cg_mg():
lx = 70
ly = 100
data, labels = make_2d_syntheticdata(lx, ly)
anticipated_warnings = [
'scipy.sparse.sparsetools|%s' % PYAMG_OR_SCIPY_WARNING,
NUMPY_MATRIX_WARNING
]
with expected_warnings(anticipated_warnings):
labels_cg_mg = random_walker(data, labels, beta=90, mode='cg_mg')
assert (labels_cg_mg[25:45, 40:60] == 2).all()
assert data.shape == labels.shape
with expected_warnings(anticipated_warnings):
full_prob = random_walker(data,
labels,
beta=90,
mode='cg_mg',
return_full_prob=True)
assert (full_prob[1, 25:45, 40:60] >= full_prob[0, 25:45, 40:60]).all()
assert data.shape == labels.shape
return data, labels_cg_mg
def run(self, ips, snap, img, para=None):
msk = np.zeros_like(img)
msk[img > para['thr2']] = 1
msk[img < para['thr1']] = 2
msk = random_walker(snap,
msk,
beta=para['beta'],
mode=para['mode'],
tol=para['tol']) == 1
ips.lut = self.buflut
if para['out'] == 'mask': img[~msk], img[msk] = ips.range
else: img[binary_dilation(msk) ^ msk] = ips.range[1]
def RandomWalker(data, marker_mask):
data = rescale_intensity(data.astype('float'), in_range=(0,255), out_range=(0,1))
data = rescale_intensity(data, in_range=(-0.35, 1 + 0.35), out_range=(-1, 1))
# The range of the binary image spans over (-1, 1).
# We choose the hottest and the coldest pixels as markers.
markers = np.zeros(data.shape, dtype=np.uint)
markers[data < -0.95] = 1;
markers[data > 0.95] = 2
# Run random walker algorithm
labels = random_walker(data, markers, beta=10, mode='bf')
def compute_cell_bmap(cell_probs):
output = cell_probs
output2 = output[::2, ::2]
local_maxi = peak_local_max(output2, indices=False, min_distance=5)
markers = ndi.label(local_maxi)[0]
markers[output2 < 0.01] = -1
segments = random_walker(output2, markers, tol=0.01)
segments = resize(segments, output.shape, order=0, preserve_range=True)
#segments = watershed(-output, markers)
gx = convolve2d(segments, np.array([[1, 0, -1]]), mode='same')
gx[0, :] = 0
gx[-1, :] = 0
gx[:, 0] = 0
gx[:, -1] = 0
gy = convolve2d(segments, np.array([[1, 0, -1]]).T, mode='same')
gy[0, :] = 0
gy[-1, :] = 0
gy[:, 0] = 0
gy[:, -1] = 0
gmag = np.sqrt(gx**2 + gy**2)
gmag = gmag > 0
D = {}
P = {}
y, x = np.where(gmag)
for i in range(y.size):
nearby_labels = np.unique(segments[y[i] - 1:y[i] + 2,
x[i] - 1:x[i] + 2])
t = tuple(nearby_labels)
if t in D.keys():
D[t].append([y[i], x[i]])
P[t].append(
np.min(cell_probs[y[i] - 1:y[i] + 2, x[i] - 1:x[i] + 2]))
else:
D[t] = [[y[i], x[i]]]
P[t] = [np.min(cell_probs[y[i] - 1:y[i] + 2, x[i] - 1:x[i] + 2])]
bmap = np.zeros(cell_probs.shape)
for t in D.keys():
coords = np.array(D[t])
#if 2-way boundary:
if len(t) < 3:
score = np.mean(np.array(P[t]))
else:
perms = permutations(t, 2)
perms = [np.mean(P[t]) for t in perms if t in P.keys()]
score = np.min(perms)
bmap[coords[:, 0], coords[:, 1]] = 1 - score
bmap[0, :] = 1
bmap[-1, :] = 1
bmap[:, 0] = 1
bmap[:, -1] = 1
return bmap
def test_3d_inactive():
n = 30
lx, ly, lz = n, n, n
data, labels = make_3d_syntheticdata(lx, ly, lz)
old_labels = np.copy(labels)
labels[5:25, 26:29, 26:29] = -1
after_labels = np.copy(labels)
with expected_warnings(['"cg" mode|CObject type' + '|' + SCIPY_EXPECTED]):
labels = random_walker(data, labels, mode='cg')
assert (labels.reshape(data.shape)[13:17, 13:17, 13:17] == 2).all()
assert data.shape == labels.shape
return data, labels, old_labels, after_labels
def test_isolated_area():
np.random.seed(0)
a = np.random.random((7, 7))
mask = -np.ones(a.shape)
# This pixel is an isolated seed
mask[1, 1] = 0
# Unlabeled pixels
mask[3:, 3:] = 0
# Seeds connected to unlabeled pixels
mask[4, 4] = 2
mask[6, 6] = 1
# Test that no error is raised, and that labels of isolated seeds are OK
with expected_warnings(
[NUMPY_MATRIX_WARNING, 'The probability range is outside']):
res = random_walker(a, mask)
assert res[1, 1] == 0
with expected_warnings(
[NUMPY_MATRIX_WARNING, 'The probability range is outside']):
res = random_walker(a, mask, return_full_prob=True)
assert res[0, 1, 1] == 0
assert res[1, 1, 1] == 0
def segment(img, markers):
"""Segment image."""
logging.info('Segmenting: %s', array_info(img))
logging.info('...with markers: %s', array_info(markers))
d = dict(
# beta=10, # Default is 130.
mode='cg_mg',
# mode=None,
multichannel=True,
spacing=img.spacing,
)
labels = segmentation.random_walker(img, markers, **d)
return labels
def robust_binarize(base_image,
_dilation=0,
heterogeity_size=10,
feature_size=50):
"""
Robust binarization algorithm based off random walker clustering
:param base_image:
:param _dilation: if set to anything other than 0, would perform a morphological dilation using this parameter value as size
:param heterogeity_size: size of the feature (px) that the method will try to eliminate by smoothing
:param feature_size: size of the feature (px) that the method will try to segment out
:return: binary_labels
"""
if np.percentile(base_image, 99) < 0.20:
if np.percentile(base_image, 99) > 0:
mult = 0.20 / np.percentile(
base_image, 99) # poissonean background assumptions
else:
mult = 1000. / np.sum(base_image)
base_image = base_image * mult
base_image[base_image > 1] = 1
clustering_markers = np.zeros(base_image.shape, dtype=np.uint8)
selem = disk(heterogeity_size)
smooth = gaussian_filter(base_image, heterogeity_size, mode='constant')
smooth_median = median(smooth, selem)
uniform_median = median(base_image, selem)
selem2 = disk(feature_size)
local_otsu = rank.otsu(smooth_median, selem2)
uniform_median_otsu = rank.otsu(uniform_median, selem2)
clustering_markers[smooth_median < local_otsu * 0.9] = 1
clustering_markers[smooth_median > local_otsu * 1.1] = 2
# dbg.random_walker_debug(smooth_median, clustering_markers)
# dbg.robust_binarize_debug(base_image, smooth_median, smooth_median, local_otsu, clustering_markers,
# 0, uniform_median, uniform_median_otsu)
binary_labels = random_walker(
smooth_median, clustering_markers, beta=10, mode='bf') - 1
if _dilation:
selem = disk(_dilation)
binary_labels = dilation(binary_labels, selem)
# dbg.robust_binarize_debug(binary_labels, base_image)
# dbg.voronoi_debug(binary_labels, local_maxi, dist, segmented_cells_labels)
# dbg.Kristen_robust_binarize(binary_labels, base_image)
return binary_labels
def test_trivial_cases():
# When all voxels are labeled
img = np.ones((10, 10))
labels = np.ones((10, 10))
with expected_warnings(["Returning provided labels"]):
pass_through = random_walker(img, labels)
np.testing.assert_array_equal(pass_through, labels)
# When all voxels are labeled AND return_full_prob is True
labels[:, :5] = 3
expected = np.concatenate(
((labels == 1)[..., np.newaxis], (labels == 3)[..., np.newaxis]),
axis=2)
with expected_warnings(["Returning provided labels"]):
test = random_walker(img, labels, return_full_prob=True)
np.testing.assert_array_equal(test, expected)
# Unlabeled voxels not connected to seed, so nothing can be done
img = np.full((10, 10), False)
object_A = np.array([(6, 7), (6, 8), (7, 7), (7, 8)])
object_B = np.array([(3, 1), (4, 1), (2, 2), (3, 2), (4, 2), (2, 3),
(3, 3)])
for x, y in np.vstack((object_A, object_B)):
img[y][x] = True
markers = np.zeros((10, 10), dtype=np.int8)
for x, y in object_B:
markers[y][x] = 1
markers[img == 0] = -1
with expected_warnings(["All unlabeled pixels are isolated"]):
output_labels = random_walker(img, markers)
assert np.all(output_labels[markers == 1] == 1)
# Here 0-labeled pixels could not be determined (no connexion to seed)
assert np.all(output_labels[markers == 0] == -1)
with expected_warnings(["All unlabeled pixels are isolated"]):
test = random_walker(img, markers, return_full_prob=True)
def randomwalk_seg(rgb_img, mask, safe_zone_len, beta, num_points):
start = time.time()
se = np.ones((safe_zone_len, safe_zone_len))
rand_fore_points = morphology.binary_erosion(mask, se)
rand_back_points = morphology.binary_erosion(mask == 0, se)
if num_points < 0:
labels = rand_fore_points * 1.0 + 2.0 * rand_back_points
random_walker_mask = random_walker(rgb_img,
labels,
beta=beta,
multichannel=True) == 1
else:
#sampling point
fore_points = segh.get_sample_point(rgb_img, num_points,
grow_fore_points)
back_points = segh.get_sample_point(rgb_img, num_points,
grow_back_points)
labels = np.zeros((m, n))
for p in fore_points:
i, j = p[-2], p[-1]
labels[i, j] = 1
for p in back_points:
i, j = p[-2], p[-1]
labels[i, j] = 2
random_walker_mask = random_walker(rgb_img,
labels,
beta=beta,
multichannel=True) == 1
print('size ', random_walker_mask.shape, ' Random Walk processing time ',
time.time() - start)
return random_walker_mask
def test_2d_bf():
lx = 70
ly = 100
data, labels = make_2d_syntheticdata(lx, ly)
with expected_warnings([NUMPY_MATRIX_WARNING]):
labels_bf = random_walker(data, labels, beta=90, mode='bf')
assert (labels_bf[25:45, 40:60] == 2).all()
assert data.shape == labels.shape
with expected_warnings([NUMPY_MATRIX_WARNING]):
full_prob_bf = random_walker(data, labels, beta=90, mode='bf',
return_full_prob=True)
assert (full_prob_bf[1, 25:45, 40:60] >=
full_prob_bf[0, 25:45, 40:60]).all()
assert data.shape == labels.shape
# Now test with more than two labels
labels[55, 80] = 3
with expected_warnings([NUMPY_MATRIX_WARNING]):
full_prob_bf = random_walker(data, labels, beta=90, mode='bf',
return_full_prob=True)
assert (full_prob_bf[1, 25:45, 40:60] >=
full_prob_bf[0, 25:45, 40:60]).all()
assert len(full_prob_bf) == 3
assert data.shape == labels.shape
def test_2d_bf():
lx = 70
ly = 100
data, labels = make_2d_syntheticdata(lx, ly)
labels_bf = random_walker(data, labels, beta=90, mode='bf')
assert (labels_bf[25:45, 40:60] == 2).all()
full_prob_bf = random_walker(data,
labels,
beta=90,
mode='bf',
return_full_prob=True)
assert (full_prob_bf[1, 25:45, 40:60] >= full_prob_bf[0, 25:45,
40:60]).all()
# Now test with more than two labels
labels[55, 80] = 3
full_prob_bf = random_walker(data,
labels,
beta=90,
mode='bf',
return_full_prob=True)
assert (full_prob_bf[1, 25:45, 40:60] >= full_prob_bf[0, 25:45,
40:60]).all()
assert len(full_prob_bf) == 3
def apply_random_walker(image):
upper_bound = np.max(image)
img_as_float = image / upper_bound
img_as_float *= 2
img_as_float -= 1
# print(img_as_float.max(), img_as_float.min())
# print(image.min(), image.max())
markers = np.zeros(img_as_float.shape, dtype=np.uint)
markers[img_as_float < -0.90] = 1
markers[img_as_float > 0.90] = 2
return random_walker(img_as_float, markers, beta=10, mode='bf')
def seg(img):
# xg = img[:,:,1]
xg = np.copy(img)
cv2.imshow('img', xg)
cv2.waitKey()
cv2.destroyAllWindows()
a, b = np.histogram(img.flatten(), bins=30, range=[15, 255])
idx = a.argmax()
range_S = b[idx]
range_B = b[-1]
# range_S_2 = b[-5]
# range_B_2 = b[-1]
tmp = np.zeros_like(img)
tmp[(img >= range_S) * (img <= range_B)] = 255
# tmp[(img >= range_S_2) *(img <= range_B_2)] = 255
cv2.imshow('test', tmp)
cv2.waitKey()
cv2.destroyAllWindows()
m1 = xg.mean()
mask = np.zeros_like(xg)
mask[xg > m1] = 1 #
cv2.imshow('img', mask)
cv2.waitKey()
cv2.destroyAllWindows()
canny = cv2.Canny(img, 20, 60) * mask
cv2.imshow('canny', canny)
cv2.waitKey()
cv2.destroyAllWindows()
closed = np.zeros_like(canny)
closed[canny > 0] = 1
closed[mask == 0] = 2
closed[(img >= range_S) * (img <= range_B)] = 2
image_segmented = random_walker(xg, closed, beta=20, mode='bf') #, beta=1
image_segmented[image_segmented == 2] = 0
image_segmented = image_segmented.astype(np.uint8) * 255
# closed = cv2.erode(image_segmented, None, iterations=2)
cv2.imshow('img res', image_segmented)
cv2.waitKey()
cv2.destroyAllWindows()
return image_segmented
def test_multispectral_2d(dtype, channel_axis):
lx, ly = 70, 100
data, labels = make_2d_syntheticdata(lx, ly)
data = data.astype(dtype, copy=False)
data = data[..., np.newaxis].repeat(2, axis=-1) # Expect identical output
data = np.moveaxis(data, -1, channel_axis)
with expected_warnings([
'"cg" mode' + '|' + SCIPY_RANK_WARNING, NUMPY_MATRIX_WARNING,
'The probability range is outside'
]):
multi_labels = random_walker(data,
labels,
mode='cg',
channel_axis=channel_axis)
data = np.moveaxis(data, channel_axis, -1)
assert data[..., 0].shape == labels.shape
with expected_warnings(
['"cg" mode' + '|' + SCIPY_RANK_WARNING, NUMPY_MATRIX_WARNING]):
single_labels = random_walker(data[..., 0], labels, mode='cg')
assert (multi_labels.reshape(labels.shape)[25:45, 40:60] == 2).all()
assert data[..., 0].shape == labels.shape
return data, multi_labels, single_labels, labels
def draw(self, event):
x, y = event.xdata, event.ydata
x, y = int(x), int(y)
# print('x=%1.2f, y=%1.2f' % (x, y))
imd = np.sqrt((self.xx-y)**2 + (self.yy-x)**2) <= self.r
if event.button == 1:
self.im_markers[imd] = 1
elif event.button == 3:
self.im_markers[imd] = 2
elif event.button == 2:
self.im_labels = random_walker(self.im, self.im_markers, beta=10, mode='cg_mg') # 'bf'
self.ax3.imshow(self.im_labels, cmap='gray', interpolation='nearest')
self.ax2.imshow(self.im_markers, cmap='magma', interpolation='nearest')
plt.draw()
def test_walker_binary():
unet, mrcnn, annot = get_images()
annot_mask = np.where(annot > 0, 1, 0)
unet_mask = np.where(unet > 0, 1, 0)
mrcnn_mask = np.where(mrcnn > 0, 1, 0)
unet_seed, mrcnn_seed = erode_images(unet, mrcnn, 6, 6)
labels = np.where(unet_seed + mrcnn_seed > 0, 1, 0)
labels[unet_mask + mrcnn_mask == 0] = 2
data = np.dstack((unet, mrcnn))
final = random_walker(data, labels, beta=30, mode='bf')
final[final == 2] = 0
imwrite('unet.png', unet_mask * 255)
imwrite('mrcnn.png', mrcnn_mask * 255)
imwrite('final_.png', final * 255)
imwrite('annot.png', annot_mask * 255)
def produceRandomWalkerSegmentationNoContrast(filename):
image = utils.readImageGrayscaled(filename)
utils.saveImageSegment(filename, 'original_gs', image)
markers = np.zeros(image.shape, dtype=np.uint8)
markers[image < 0.25] = 2
markers[image > 0.5] = 1
image_segmented = seg.random_walker(image, markers)
utils.saveImageSegment(filename, 'rand_walk_unsupervised', image_segmented)
merged = image + image_segmented
utils.saveImageSegment(filename, 'merge_rand_walk_unsupervised', merged)
utils.saveImageSegmentResult(filename, 'rand_walk_unsupervised', merged)
def get_labels(prediction):
def _get_reduced_maxima(pred):
markers = []
distance = ndi.distance_transform_edt(pred)
distance_ordinal = distance - np.min(distance)
distance_ordinal = distance_ordinal / np.max(distance_ordinal)
markers.append(ndi.label(distance_ordinal)[0])
markers.append(
ndi.label(
ndi.distance_transform_edt(distance_ordinal >= 0.1 * 1.0))[0])
# markers.append(ndi.label(ndi.distance_transform_edt(distance_ordinal >= 0.5 * 1.0))[0])
# markers.append(ndi.label(ndi.distance_transform_edt(distance_ordinal >= 0.75 * 1.0))[0])
final_markers = ndi.label(
ndi.distance_transform_edt(distance_ordinal >= 0.5 * 1.0))[0]
# # Calculate mean cell size
cell_sums = []
for c_id in np.unique(final_markers)[1:]:
cell_sums.append(np.sum(c_id == final_markers * 1.0))
mean_sum = np.mean(cell_sums)
# Reduce centroid size
for marker in markers[::-1]:
transition_markers = marker * (final_markers > 0)
for c_id in np.unique(marker)[1:]:
cell_seg = (marker == c_id) * 1.0
if np.sum(cell_seg *
(transition_markers > 0 * 1.0)) <= mean_sum * 0.3:
final_markers = ndi.label((final_markers > 0) * 1.0 +
(marker == c_id) * 1.0)[0]
return final_markers, mean_sum
pred, ind = get_expanded_boundary(prediction)
markers, mean_sum = _get_reduced_maxima(pred)
markers = get_normal_boundary(markers, ind)
markers = markers * (prediction > 0 * 1.0) - 1 * (prediction == 0 * 1.0)
wa = random_walker(prediction, markers)
wa = wa * (wa != -1)
# Remove small cell artifacts
wa_2 = np.zeros(np.shape(wa))
for c_id in np.unique(wa)[1:]:
if np.sum((c_id == wa) * 1.0) >= mean_sum * 0.1:
wa_2 += (c_id == wa) * c_id
wa = wa_2
return wa
def get_regions(img):
denoised=restoration.denoise_bilateral(img.astype('uint16'),
# sigma_range=0.01,
sigma_spatial=15,
multichannel=False)
smoothened = filters.median(denoised,np.ones((4,4)))
markers = np.zeros(smoothened.shape, dtype=np.uint)
# otsu works only for only for multi-channel images
# markers[smoothened < filters.threshold_otsu(smoothened)] = 1
# markers[smoothened > filters.threshold_otsu(smoothened)] = 2
markers[smoothened < filters.median(smoothened)] = 1
markers[smoothened > filters.median(smoothened)] = 2
labels = random_walker(smoothened, markers, beta=10, mode='bf')
regions= measure.label(labels)
return regions, denoised, smoothened,markers
def random_walker_binarize(base_image, _dilation=0):
gfp_clustering_markers = np.zeros(base_image.shape, dtype=np.uint8)
# To try: add a grey area around the boundary between black and white
gfp_clustering_markers[base_image > np.mean(base_image) * 2] = 2
gfp_clustering_markers[base_image < np.mean(base_image) * 0.20] = 1
binary_labels = random_walker(
base_image, gfp_clustering_markers, beta=10, mode='bf') - 1
if _dilation:
selem = disk(_dilation)
binary_labels = dilation(binary_labels, selem)
return binary_labels
def label_particles_walker(im, min_thresh=0.3, max_thresh=0.5, sigma=3):
""" label_particles_walker(image, min_thresh=0.3, max_thresh=0.5)
Returns the labels for an image.
Segments using random_walker method.
keyword arguments:
image -- The image in which to find particles
min_thresh -- The lower limit for binary threshold
max_thresh -- The upper limit for binary threshold
"""
if sigma > 0:
im = gaussian_filter(im, sigma)
labels = np.zeros_like(im)
labels[im < min_thresh * im.max()] = 1
labels[im > max_thresh * im.max()] = 2
return segmentation.random_walker(im, labels)
def background_generator():
size_x = 50
size_y = 50
n_labels = 2
image_rgb = np.random.uniform(0.9, 1, (size_x, size_y, 3))
labels = np.random.randint(n_labels + 1, size=(
size_x, size_y)) * np.random.randint(0, 2, size=(size_x, size_y))
# segment random image
segments = random_walker(image_rgb,
labels,
multichannel=True,
beta=250,
copy=False,
spacing=[50, 10])
for color_index in np.unique(segments):
image_rgb[segments == color_index] = np.random.uniform(0, 1, 3)
# transform segmented image so it is large, preserving blobs, and blurry
image_rgb = rescale(image_rgb, 2, anti_aliasing=False, multichannel=True)
image_rgb = gaussian(image_rgb, sigma=2, multichannel=True)
image_hsv = color.rgb2hsv(image_rgb)
buffer = 0.05
hsv_index = 1
for _ in range(10000):
rand_x, rand_y = np.random.choice(
image_hsv.shape[0]), np.random.choice(image_hsv.shape[1])
orig_rgb = image_rgb[rand_x, rand_y]
rand_value = image_hsv[rand_x, rand_y, hsv_index]
x, y = np.where((rand_value *
(1 + buffer) >= image_hsv[:, :, hsv_index])
& (rand_value *
(1 - buffer) < image_hsv[:, :, hsv_index]))
if len(x) == 0:
continue
idx = np.random.choice(len(x))
update_rgb = image_rgb[x[idx], y[idx]]
image_rgb[x[idx], y[idx]] = orig_rgb
image_rgb[rand_x, rand_y] = update_rgb
return image_rgb
def watershedPower(input_file, output_file, seeds_file, algo, mult, geod):
image_name = "figure_1.png"
seeds_name = seeds_file
sigma = 0.35
fimage_read = open(image_name, 'rb')
fimage_seeds = open(seeds_name, 'rb')
image_read = read_pgm(fimage_read)
seeds = read_pgm(fimage_seeds)
rs = rowsize(fimage_read)
cs = colsize(fimage_read)
N = rs * cs
M = rs * (cs - 1) + (rs - 1) * cs #number of edges
nblabels = 2
j = 0
for i in range(0, rs * cs):
if seeds[i] > 155:
index_seeds[j] = i
index_labels[j] = 1
j += 1
if seeds[i] < 100:
index_seeds[j] = i
index_labels[j] = 2
j += 1
size_seeds = j
markers = np.zeros(image.shape, dtype=np.uint)
markers[image < -0.95] = 1
markers[image > 0.95] = 2
matrixedges = [[0 for col in range(2)]
for row in range((cs - 1) * rs + (rs - 1) * cs)]
edges = compute_edges(matrixedges, rs, cs)
weights = []
normal_weights = grey_weights_PW(image_name, edges, index_seeds,
size_seeds, weights, quicksort)
output = PowerWatershed_q2(edges, weights, normal_weights, max_weight,
index_seeds, index_labels, size_seeds, rs, cs,
ds, nblabels, quicksort, img_proba)
labels = random_walker(image, markers, beta=10, mode='bf')
def test_walker_label():
unet, mrcnn, annot = get_images()
unet_seed, mrcnn_seed = erode_images(unet, mrcnn, 6, 6)
annot_mask = np.where(annot > 0, 1, 0)
unet_mask = np.where(unet > 0, 1, 0)
mrcnn_mask = np.where(mrcnn > 0, 1, 0)
labels = np.where(unet_seed + mrcnn_seed > 0, 1, 0)
labels = label(labels)
labels, _, _ = relabel_sequential(labels, offset=2)
labels[unet_mask + mrcnn_mask == 0] = 1
data = np.dstack((unet, mrcnn))
final = random_walker(data, labels, beta=50, mode='bf')
final[final == 1] = 0
imwrite('unet.png', unet_mask*255)
imwrite('mrcnn.png', mrcnn_mask*255)
imwrite('final_.png', final*255)
imwrite('annot.png', annot_mask * 255)
return final
def add_missed_blobs(full_mask, labeled_mask, edges):
missed_mask = full_mask & ~(labeled_mask > 0)
missed_mask = drop_small_blobs(missed_mask,
2) # bodies must be larger than 1-pixel
if edges is not None:
missed_markers = label(missed_mask & ~edges)
else:
missed_markers = label(missed_mask)
if missed_markers.max() > 0:
missed_markers[missed_mask == 0] = -1
missed_labels = random_walker(missed_mask, missed_markers)
missed_labels[missed_labels <= 0] = 0
missed_labels = np.where(missed_labels > 0,
missed_labels + labeled_mask.max(), 0)
final_labels = np.add(labeled_mask, missed_labels)
else:
final_labels = labeled_mask
return final_labels
def rondomwalker_seg(distance, threshold=2.0):
if isinstance(distance, Variable):
distance = distance.data
if isinstance(distance, torch.FloatTensor) or isinstance(
distance, torch.cuda.FloatTensor):
distance = distance.cpu().numpy()
#print(type(distance))
distance = np.squeeze(distance)
distance = gaussian(distance / np.max(distance), sigma=0.6, mode='reflect')
markers = distance > 0.039
markers = skimage.morphology.label(markers)
seg_labels = random_walker(-distance, markers,\
beta=25000, mode='cg_mg')
#seg_labels = watershed(-distance, markers)
#axes[1].plot(coordinates[:, 1], coordinates[:, 0], 'r.')
return seg_labels
def img_random_walker(data,
lower_q=0.9,
upper_q=0.98,
val_beta=10,
val_mode='bf'):
labels = []
if not isinstance(data, list):
raise Exception('Argument data must be a list type')
for x in data:
ones = np.ones(x.shape)
markers = np.zeros(x.shape)
markers[x < np.quantile(x, lower_q)] = 1
markers[x > np.quantile(x, upper_q)] = 2
labels.append(
random_walker(x, markers, beta=val_beta, mode=val_mode) - ones)
return labels
