def peak_finder(X, thresh):
"""Simple peak finding algorithm.
Parameters
----------
X : array_like, shape (n_times,)
Raw data trace.
thresh : float
Amplitude threshold.
Returns
-------
peak_loc : array_like, shape (n_clusters,)
Index of peak amplitudes.
peak_mag : array_like, shape (n_clusters,)
Magnitude of peak amplitudes.
"""
## Error-catching.
assert X.ndim == 1
## Identify clusters.
clusters, ix = measurements.label(X > thresh)
## Identify index of peak amplitudes.
peak_loc = np.concatenate(
measurements.maximum_position(X,
labels=clusters,
index=np.arange(ix) + 1))
## Identify magnitude of peak amplitudes.
peak_mag = measurements.maximum(X,
labels=clusters,
index=np.arange(ix) + 1)
return peak_loc, peak_mag
def peak_finder(arr, thresh):
'''Absolute amplitude-based peak finding algorithm.
Parameters
----------
arr : array, shape=(n_samples,)
1-d timeseries of extracellular recording.
thresh : scalar
amplitude threshold. Only samples above this
value will be included in clusters.
Returns
-------
peak_loc : array
indices of cluster peaks.
peak_mag : array
magnitude of cluster peaks.
'''
assert arr.ndim == 1
clusters, ix = measurements.label(arr > thresh)
if not ix: return np.array([]), np.array([])
peak_loc = np.concatenate(
measurements.maximum_position(arr,
labels=clusters,
index=np.arange(ix) + 1))
peak_mag = measurements.maximum(arr,
labels=clusters,
index=np.arange(ix) + 1)
return peak_loc, peak_mag
def find_blob_centers(predictions, resolution, blob_prediction_threshold,
blob_size_threshold):
# smooth out "U-Net noise"
# print("Median-filtering prediction...")
# predictions = median_filter(predictions, size=3)
print("Finding blobs...")
start = time.time()
blobs = predictions > blob_prediction_threshold
labels, num_blobs = label(blobs)
print("%.3fs" % (time.time() - start))
print("Found %d blobs" % num_blobs)
print("Finding centers, sizes, and maximal values...")
start = time.time()
label_ids = np.arange(1, num_blobs + 1)
centers = measurements.center_of_mass(blobs, labels, index=label_ids)
sizes = measurements.sum(blobs, labels, index=label_ids)
maxima = measurements.maximum(predictions, labels, index=label_ids)
print("%.3fs" % (time.time() - start))
centers = {
label: {
'center': center,
'score': max_value
}
for label, center, size, max_value in zip(
label_ids, centers, sizes, maxima) if size >= blob_size_threshold
}
return (centers, labels)
def find_maxima(predictions,
voxel_size,
radius,
sigma=None,
min_score_threshold=0):
'''Find all points that are maximal within a sphere of ``radius`` and are
strictly higher than min_score_threshold. Optionally smooth the prediction
with sigma.'''
# smooth predictions
if sigma is not None:
print("Smoothing predictions...")
sigma = tuple(float(s) / r for s, r in zip(sigma, voxel_size))
print("voxel-sigma: %s" % (sigma, ))
start = time.time()
predictions = gaussian_filter(predictions, sigma, mode='constant')
print("%.3fs" % (time.time() - start))
print("Finding maxima...")
start = time.time()
radius = tuple(
int(math.ceil(float(ra) / re)) for ra, re in zip(radius, voxel_size))
print("voxel-radius: %s" % (radius, ))
max_filtered = maximum_filter(predictions, footprint=sphere(radius))
maxima = max_filtered == predictions
print("%.3fs" % (time.time() - start))
print("Applying NMS...")
start = time.time()
predictions_filtered = np.zeros_like(predictions)
predictions_filtered[maxima] = predictions[maxima]
print("%.3fs" % (time.time() - start))
print("Finding blobs...")
start = time.time()
blobs = predictions_filtered > min_score_threshold
labels, num_blobs = label(blobs, output=np.uint64)
print("%.3fs" % (time.time() - start))
print("Found %d points after NMS" % num_blobs)
print("Finding centers, sizes, and maximal values...")
start = time.time()
label_ids = np.arange(1, num_blobs + 1)
centers = measurements.center_of_mass(blobs, labels, index=label_ids)
sizes = measurements.sum(blobs, labels, index=label_ids)
maxima = measurements.maximum(predictions, labels, index=label_ids)
print("%.3fs" % (time.time() - start))
centers = {
label: {
'center': center,
'score': max_value
}
for label, center, size, max_value in zip(label_ids, centers, sizes,
maxima)
}
return (centers, labels, predictions)
def measure(im, labels, num_objs, measurement_type='mean'):
if measurement_type == 'mean':
res = spm.mean(im, labels, range(1, num_objs))
elif measurement_type == 'max':
res = spm.maximum(im, labels, range(1, num_objs))
else:
raise ValueError('Unsporrted measurement type: {}'.format(measurement_type))
return res
def amplitude_thresholding(x0, thresh):
'''Simple peak finding algorithm.'''
assert x0.ndim == 1
clusters, ix = measurements.label(x0 > thresh)
peak_loc = np.concatenate(
measurements.maximum_position(x0,
labels=clusters,
index=np.arange(ix) + 1))
peak_mag = measurements.maximum(x0,
labels=clusters,
index=np.arange(ix) + 1)
return peak_loc, peak_mag
def _localize_process(data: tuple,
is_binary: bool = True,
use_labels: bool = False) -> np.ndarray:
# image: np.ndarray, frame: int) -> np.ndarray:
""" worker process for localizing and labelling objects
volumetric data is usually of the format: t, z, x, y
Returns:
combined data in form of nx5 array (t, x, y, z, label) adding a
z-dimension of uniform zero, if one doesn't exist.
"""
if use_labels: assert is_binary
image, frame = data
assert image.dtype in (np.uint8, np.uint16)
if is_binary:
labeled, n = measurements.label(image.astype(np.bool))
idx = list(range(1, n + 1))
else:
labeled = image
idx = [p for p in np.unique(labeled) if p > 0]
# calculate the centroids
centroids = np.array(
measurements.center_of_mass(image, labels=labeled, index=idx))
# if we're dealing with volumetric data, reorder so that z is last
if image.ndim == 3:
centroids = np.roll(centroids, -1)
localizations = np.zeros((centroids.shape[0], 5), dtype=np.uint16)
localizations[:, 0] = frame # time
localizations[:, 1:centroids.shape[1] + 1] = centroids
localizations[:,
-1] = 0 #labels-1 #-1 because we use a label of zero for states
# if we're not using labels from the segmentation, return here
if not use_labels:
return localizations
# get the labels from the image data
labels = np.array(measurements.maximum(image, labels=labeled, index=idx))
localizations[:,
-1] = labels - 1 #-1 because we use a label of zero for states
return localizations
def local_errors(self, threshold):
subregion = crop_region(patch_size, ERRORS_CROP)
unique_list = unique_nonzero(self.raw_labels[subregion])
max_error_list = measurements.maximum(self.errors, self.raw_labels,
unique_list)
additional_segments = [
unique_list[i] for i in xrange(len(unique_list))
if max_error_list[i] > threshold or max_error_list[i] == 0.0
]
additional_segments = filter(
lambda x: x != 0 and x not in self.parent.valid,
additional_segments)
return additional_segments
def dsm(parameters):
"""Generates a DSM by elevating groups a cells by certain height.
This requires an input array, the DTM, and a mask. The mask designates
which cells of the DTM should be elevated in order to produce the DSM.
Basically, the mask shows in which cells there are features with
significant height, e.g. trees, buildings etc.
The tricky part it to account for DTM slope when elevating a group of
cells. If you simply add some height to the initial DTM then the features
will be elevated parallel to the ground. Especially in the case of
buildings, their roof is horizontal, regardless of the underlying DTM
slope.
To account for this, the algorithm initially labels the mask. As a result
you get groups of cells which should all be elevated to the same height.
Next, it finds the maximum height value of underlying DTM for each blob.
Finally, it assigns `max_blob_height + delta_height` to each blob cell.
:param parameters['data'][0]: the base DTM
:type parameters['data'][0]: numpy.array
:param parameters['data'][1]: the mask of cells to elevate
:type parameters['data'][1]: numpy.array with boolean/binary values
:param parameters['delta_height']: single cell elevation value
:type parameters['delta_height']: float or integer
:return: numpy.array
"""
dtm = parameters['data'][0]
mask = parameters['data'][1]
delta_height = parameters['delta_height']
# label and find the max height of each blob
labels, count = measurements.label(mask)
max_heights = measurements.maximum(dtm,
labels=labels,
index=range(1, count + 1))
# assign the max height at each blob cell, required to copy so it won't
# change the initial dtm values
dsm = dtm.copy()
for blob_id in range(1, count + 1):
dsm[numpy.where(labels == blob_id)] = max_heights[blob_id - 1] +\
delta_height
return dsm
def detect(self):
"""
a method for detecting peaks, trying to speed things up!
Run this after running the thresholding (which sets the pk_mask)!
"""
self.pos = []
self.intens = []
sz = self.pk_par["sz"]
sig_G = self.pk_par["sig_G"]
min_conn = self.pk_par["min_conn"]
max_conn = self.pk_par["max_conn"]
peak_COM = self.pk_par["peak_COM"]
self.lab_img, _ = measurements.label(self.img * self.pk_mask)
obs = measurements.find_objects(self.lab_img)
for i, (sy, sx) in enumerate(obs):
lab_idx = i + 1
y1 = max(0, sy.start - sz)
y2 = min(self.img_sh[0], sy.stop + sz)
x1 = max(0, sx.start - sz)
x2 = min(self.img_sh[1], sx.stop + sz)
l = self.lab_img[y1:y2, x1:x2]
nconn = np.sum(l == (lab_idx))
if nconn < min_conn:
continue
if nconn > max_conn:
continue
pix = self.img[y1:y2, x1:x2]
self.intens.append(measurements.maximum(pix, l, lab_idx))
if peak_COM:
y, x = measurements.center_of_mass(pix, l, lab_idx)
else:
y, x = measurements.maximum_position(pix, l, lab_idx)
self.pos.append((y + y1, x + x1))
self.intens = np.array(self.intens)
def find_max_points(
predictions,
resolution,
radius,
sigma=None,
min_score_threshold=0):
'''Find all points that are maximal withing a sphere of ``radius`` and are
strictly higher than min_score_threshold. Optionally smooth the prediction
with sigma.'''
# smooth predictions
if sigma is not None:
predictions = smooth(predictions, resolution, sigma)
maxima = find_maxima(predictions, resolution, radius)
predictions_filtered = apply_nms(predictions, maxima)
print("Finding blobs...")
start = time.time()
blobs = predictions_filtered > min_score_threshold
labels, num_blobs = label(blobs)
print("%.3fs"%(time.time()-start))
print("Found %d blobs after NMS"%num_blobs)
print("Finding centers, sizes, and maximal values...")
start = time.time()
label_ids = np.arange(1, num_blobs + 1)
centers = measurements.center_of_mass(blobs, labels, index=label_ids)
sizes = measurements.sum(blobs, labels, index=label_ids)
maxima = measurements.maximum(predictions, labels, index=label_ids)
print("%.3fs"%(time.time()-start))
centers = {
str(label): { 'center': center, 'score': max_value }
for label, center, size, max_value in zip(label_ids, centers, sizes, maxima)
}
return (centers, labels)
def roi_max_counts(images_sets, label_array):
"""
Return the brightest pixel in any ROI in any image in the image set.
Parameters
----------
images_sets : array
iterable of 4D arrays
shapes is: (len(images_sets), )
label_array : array
labeled array; 0 is background.
Each ROI is represented by a distinct label (i.e., integer).
Returns
-------
max_counts : int
maximum pixel counts
"""
max_cts = 0
for img_set in images_sets:
for img in img_set:
max_cts = max(max_cts, ndim.maximum(img, label_array))
return max_cts
def roi_max_counts(images_sets, label_array):
"""
Return the brightest pixel in any ROI in any image in the image set.
Parameters
----------
images_sets : array
iterable of 4D arrays
shapes is: (len(images_sets), )
label_array : array
labeled array; 0 is background.
Each ROI is represented by a distinct label (i.e., integer).
Returns
-------
max_counts : int
maximum pixel counts
"""
max_cts = 0
for img_set in images_sets:
for img in img_set:
max_cts = max(max_cts, ndim.maximum(img, label_array))
return max_cts
def max_clustering(Response, Mask, r=10):
"""Local max clustering pixel aggregation for nuclear segmentation.
Takes as input a constrained log or other filtered nuclear image, a binary
nuclear mask, and a clustering radius. For each pixel in the nuclear mask,
the local max is identified. A hierarchy of local maxima is defined, and
the root nodes used to define the label image.
Parameters
----------
Response : array_like
A filtered-smoothed image where the maxima correspond to nuclear
center. Typically obtained by constrained-LoG filtering on a
hematoxylin intensity image obtained from ColorDeconvolution.
Mask : array_like
A binary mask of type boolean where nuclei pixels have value
'True', and non-nuclear pixels have value 'False'.
r : float
A scalar defining the clustering radius. Default value = 10.
Returns
-------
Label : array_like
Label image where positive values correspond to foreground pixels that
share mutual sinks.
Seeds : array_like
An N x 2 array defining the (x,y) coordinates of nuclei seeds.
Maxima : array_like
An N x 1 array containing the maximum response value corresponding to
'Seeds'.
See Also
--------
histomicstk.filters.shape.clog
References
----------
.. [1] XW. Wu et al "The local maximum clustering method and its
application in microarray gene expression data analysis,"
EURASIP J. Appl. Signal Processing, volume 2004, no.1, pp.53-63,
2004.
.. [2] Y. Al-Kofahi et al "Improved Automatic Detection and Segmentation
of Cell Nuclei in Histopathology Images" in IEEE Transactions on
Biomedical Engineering,vol.57,no.4,pp.847-52, 2010.
"""
# check type of input mask
if Mask.dtype != np.dtype("bool"):
raise TypeError("Input 'Mask' must be a bool")
# define kernel for max filter
Kernel = np.zeros((2 * r + 1, 2 * r + 1), dtype=bool)
X, Y = np.meshgrid(np.linspace(0, 2 * r, 2 * r + 1), np.linspace(0, 2 * r, 2 * r + 1))
X -= r
Y -= r
Kernel[(X ** 2 + Y ** 2) ** 0.5 <= r] = True
# define linear coordinates of postive kernel entries
X = X[Kernel].astype(np.int)
Y = Y[Kernel].astype(np.int)
# pad input array to simplify filtering
I = Response.min() * np.ones((Response.shape[0] + 2 * r, Response.shape[1] + 2 * r))
MaskedResponse = Response.copy()
MaskedResponse[~Mask] = Response.min()
I[r : r + Response.shape[0], r : r + Response.shape[1]] = MaskedResponse
# initialize coordinate arrays and max value arrays
Max = np.zeros(I.shape)
Row = np.zeros(I.shape, dtype=np.int)
Col = np.zeros(I.shape, dtype=np.int)
# define pixels for local neighborhoods
py, px = np.nonzero(Mask)
py = py + np.int(r)
px = px + np.int(r)
# perform max filtering
for i in np.arange(0, px.size, 1):
# calculate local max value and position around px[i], py[i]
Index = np.argmax(I[py[i] + Y, px[i] + X])
Max[py[i], px[i]] = I[py[i] + Y[Index], px[i] + X[Index]]
Row[py[i], px[i]] = py[i] + Y[Index] - r
Col[py[i], px[i]] = px[i] + X[Index] - r
# trim outputs
Max = Max[r : Response.shape[0] + r, r : Response.shape[1] + r]
Row = Row[r : Response.shape[0] + r, r : Response.shape[1] + r]
Col = Col[r : Response.shape[0] + r, r : Response.shape[1] + r]
# subtract out padding offset for px, py
py = py - r
px = px - r
# identify connected regions of local maxima and define their seeds
Label = spm.label((Response == Max) & Mask)[0]
Seeds = np.array(spm.center_of_mass(Response, Label, np.arange(1, Label.max() + 1)))
Seeds = np.round(Seeds).astype(np.uint32)
# capture maxima for each connected region
Maxima = spm.maximum(Response, Label, np.arange(1, Label.max() + 1))
# handle seeds lying outside non-convex objects
Fix = np.nonzero(
Label[Seeds[:, 0].astype(np.uint32), Seeds[:, 1].astype(np.uint32)] != np.arange(1, Label.max() + 1)
)[0]
if Fix.size > 0:
Locations = spm.find_objects(Label)
for i in np.arange(Fix.size):
Patch = Label[Locations[Fix[i]]]
Pixels = np.nonzero(Patch)
dX = Pixels[1] - (Seeds[Fix[i], 1] - Locations[Fix][1].start)
dY = Pixels[0] - (Seeds[Fix[i], 0] - Locations[Fix][0].start)
Dist = (dX ** 2 + dY ** 2) ** 0.5
NewSeed = np.argmin(Dist)
Seeds[Fix[i], 1] = np.array(Locations[Fix][1].start + Pixels[1][NewSeed]).astype(np.uint32)
Seeds[Fix[i], 0] = np.array(Locations[Fix][0].start + Pixels[0][NewSeed]).astype(np.uint32)
# initialize tracking and segmentation masks
Tracked = np.zeros(Max.shape, dtype=bool)
Tracked[Label > 0] = True
# track each pixel and update
for i in np.arange(0, px.size, 1):
# initialize tracking trajectory
Id = 0
Alloc = 1
Trajectory = np.zeros((1000, 2), dtype=np.int)
Trajectory[0, 0] = px[i]
Trajectory[0, 1] = py[i]
while ~Tracked[Trajectory[Id, 1], Trajectory[Id, 0]]:
# increment trajectory counter
Id += 1
# if overflow, copy and reallocate
if Id == 1000 * Alloc:
temp = Trajectory
Trajectory = np.zeros((1000 * (Alloc + 1), 2), dtype=np.int)
Trajectory[0 : 1000 * (Alloc),] = temp
Alloc += 1
# add local max to trajectory
Trajectory[Id, 0] = Col[Trajectory[Id - 1, 1], Trajectory[Id - 1, 0]]
Trajectory[Id, 1] = Row[Trajectory[Id - 1, 1], Trajectory[Id - 1, 0]]
# label sequence and add to tracked list
Tracked[Trajectory[0:Id, 1], Trajectory[0:Id, 0]] = True
Label[Trajectory[0:Id, 1], Trajectory[0:Id, 0]] = Label[Trajectory[Id, 1], Trajectory[Id, 0]]
# return
return Label, Seeds, Maxima
def compute_fixations(aligned, times, labels=None):
"""Compute fixations from aligned timeseries. Fixations are defined
as contiguous samples of eyetracking data aligned to the same AoI.
Parameters
----------
aligned : array, shape (n_trials, n_times)
Eyetracking timeseries aligned to areas of interest.
times : array, shape (n_times,)
Time vector in seconds.
labels : list
List of areas of interest to include in processing. Defaults
to all non-zero values in aligned.
Returns
-------
fixations : pd.DataFrame
Pandas DataFrame where each row details the (Trial, AoI,
Onset, Offset, Duration) of the fixation.
Notes
-----
Currently supports only monocular data. In the case of binocular
data, the user can simply pass the aligned object twice (once
per eye).
"""
## Error-catching.
assert np.ndim(aligned) == 2
assert np.shape(aligned)[-1] == np.size(times)
## Define labels list.
if labels is None: labels = [i for i in np.unique(aligned) if i]
## Append extra timepoint to end of each trial. This prevents clusters across
## successive trials.
n_trials, n_times = aligned.shape
aligned = np.hstack([aligned, np.zeros((n_trials, 1))]).flatten()
## Precompute trial info.
trials = np.repeat(np.arange(n_trials), n_times + 1) + 1
times = np.broadcast_to(np.append(times, 0),
(n_trials, n_times + 1)).flatten()
## Preallocate space.
df = DataFrame([], columns=('Trial', 'AoI', 'Onset', 'Offset'))
for label in labels:
## Identify clusters.
clusters, n_clusters = measurements.label(aligned == label)
## Identify cluster info.
trial = measurements.minimum(trials,
labels=clusters,
index=np.arange(n_clusters) + 1)
onset = measurements.minimum(times,
labels=clusters,
index=np.arange(n_clusters) + 1)
offset = measurements.maximum(times,
labels=clusters,
index=np.arange(n_clusters) + 1)
## Append to DataFrame.
dat = np.column_stack(
(trial, np.ones_like(trial) * label, onset, offset))
df = df.append(DataFrame(dat, columns=df.columns))
df = df.sort_values(['Trial', 'Onset']).reset_index(drop=True)
df['Duration'] = df.Offset - df.Onset
return df
def max_clustering(Response, Mask, r=10):
"""Local max clustering pixel aggregation for nuclear segmentation.
Takes as input a constrained log or other filtered nuclear image, a binary
nuclear mask, and a clustering radius. For each pixel in the nuclear mask,
the local max is identified. A hierarchy of local maxima is defined, and
the root nodes used to define the label image.
Parameters
----------
Response : array_like
A filtered-smoothed image where the maxima correspond to nuclear
center. Typically obtained by constrained-LoG filtering on a
hematoxylin intensity image obtained from ColorDeconvolution.
Mask : array_like
A binary mask of type boolean where nuclei pixels have value
'True', and non-nuclear pixels have value 'False'.
r : float
A scalar defining the clustering radius. Default value = 10.
Returns
-------
Label : array_like
Label image where positive values correspond to foreground pixels that
share mutual sinks.
Seeds : array_like
An N x 2 array defining the (x,y) coordinates of nuclei seeds.
Maxima : array_like
An N x 1 array containing the maximum response value corresponding to
'Seeds'.
See Also
--------
histomicstk.filters.shape.clog
References
----------
.. XW. Wu et al "The local maximum clustering method and its
application in microarray gene expression data analysis,"
EURASIP J. Appl. Signal Processing, volume 2004, no.1, pp.53-63,
2004.
.. Y. Al-Kofahi et al "Improved Automatic Detection and Segmentation
of Cell Nuclei in Histopathology Images" in IEEE Transactions on
Biomedical Engineering,vol.57,no.4,pp.847-52, 2010.
"""
# check type of input mask
if Mask.dtype != np.dtype('bool'):
raise TypeError("Input 'Mask' must be a bool")
# define kernel for max filter
Kernel = np.zeros((2*r+1, 2*r+1), dtype=bool)
X, Y = np.meshgrid(np.linspace(0, 2*r, 2*r+1), np.linspace(0, 2*r, 2*r+1))
X -= r
Y -= r
Kernel[(X**2 + Y**2)**0.5 <= r] = True
# define linear coordinates of postive kernel entries
X = X[Kernel].astype(np.int)
Y = Y[Kernel].astype(np.int)
# pad input array to simplify filtering
I = Response.min() * np.ones((Response.shape[0]+2*r,
Response.shape[1]+2*r))
MaskedResponse = Response.copy()
MaskedResponse[~Mask] = Response.min()
I[r:r+Response.shape[0], r:r+Response.shape[1]] = MaskedResponse
# initialize coordinate arrays and max value arrays
Max = np.zeros(I.shape)
Row = np.zeros(I.shape, dtype=np.int)
Col = np.zeros(I.shape, dtype=np.int)
# define pixels for local neighborhoods
py, px = np.nonzero(Mask)
py = py + np.int(r)
px = px + np.int(r)
# perform max filtering
for i in np.arange(0, px.size, 1):
# calculate local max value and position around px[i], py[i]
Index = np.argmax(I[py[i]+Y, px[i]+X])
Max[py[i], px[i]] = I[py[i]+Y[Index], px[i]+X[Index]]
Row[py[i], px[i]] = py[i] + Y[Index] - r
Col[py[i], px[i]] = px[i] + X[Index] - r
# trim outputs
Max = Max[r:Response.shape[0]+r, r:Response.shape[1]+r]
Row = Row[r:Response.shape[0]+r, r:Response.shape[1]+r]
Col = Col[r:Response.shape[0]+r, r:Response.shape[1]+r]
# subtract out padding offset for px, py
py = py - r
px = px - r
# identify connected regions of local maxima and define their seeds
Label = spm.label((Response == Max) & Mask)[0]
Seeds = np.array(spm.center_of_mass(Response, Label,
np.arange(1, Label.max()+1)))
Seeds = np.round(Seeds).astype(np.uint32)
# capture maxima for each connected region
Maxima = spm.maximum(Response, Label, np.arange(1, Label.max()+1))
# handle seeds lying outside non-convex objects
Fix = np.nonzero(Label[Seeds[:, 0].astype(np.uint32),
Seeds[:, 1].astype(np.uint32)] !=
np.arange(1, Label.max()+1))[0]
if(Fix.size > 0):
Locations = spm.find_objects(Label)
for i in np.arange(Fix.size):
Patch = Label[Locations[Fix[i]]]
Pixels = np.nonzero(Patch)
dX = Pixels[1] - (Seeds[Fix[i], 1] - Locations[Fix][1].start)
dY = Pixels[0] - (Seeds[Fix[i], 0] - Locations[Fix][0].start)
Dist = (dX**2 + dY**2)**0.5
NewSeed = np.argmin(Dist)
Seeds[Fix[i], 1] = np.array(Locations[Fix][1].start +
Pixels[1][NewSeed]).astype(np.uint32)
Seeds[Fix[i], 0] = np.array(Locations[Fix][0].start +
Pixels[0][NewSeed]).astype(np.uint32)
# initialize tracking and segmentation masks
Tracked = np.zeros(Max.shape, dtype=bool)
Tracked[Label > 0] = True
# track each pixel and update
for i in np.arange(0, px.size, 1):
# initialize tracking trajectory
Id = 0
Alloc = 1
Trajectory = np.zeros((1000, 2), dtype=np.int)
Trajectory[0, 0] = px[i]
Trajectory[0, 1] = py[i]
while(~Tracked[Trajectory[Id, 1], Trajectory[Id, 0]]):
# increment trajectory counter
Id += 1
# if overflow, copy and reallocate
if(Id == 1000*Alloc):
temp = Trajectory
Trajectory = np.zeros((1000*(Alloc+1), 2), dtype=np.int)
Trajectory[0:1000*(Alloc), ] = temp
Alloc += 1
# add local max to trajectory
Trajectory[Id, 0] = Col[Trajectory[Id-1, 1], Trajectory[Id-1, 0]]
Trajectory[Id, 1] = Row[Trajectory[Id-1, 1], Trajectory[Id-1, 0]]
# label sequence and add to tracked list
Tracked[Trajectory[0:Id, 1], Trajectory[0:Id, 0]] = True
Label[Trajectory[0:Id, 1], Trajectory[0:Id, 0]] = \
Label[Trajectory[Id, 1], Trajectory[Id, 0]]
# return
return Label, Seeds, Maxima
def get_feature_snr(self, segment):
return measurements.maximum(self.original_image.data, self.labels,
segment.get_segmentid()) / self.rms_noise
def get_feature_snr(self, segment):
return measurements.maximum(self.original_image.data, self.labels,
segment.get_segmentid()) / self.rms_noise
