def segment_secondary(self, marker_channel_name):
# 1. load image
img = self.load().astype(np.float64) / 255.0
mask = np.ones(img.shape)
# 2. get primary objects
labels_in = np.zeros(img.shape, dtype=float)
marker_pks = np.array([marker.pk for marker in self.gon.composite.markers.filter(channel__name=marker_channel_name, track_instance__t=self.t)])
for marker in self.gon.composite.markers.filter(channel__name=marker_channel_name, track_instance__t=self.t):
labels_in[marker.r-3:marker.r+2, marker.c-3:marker.c+2] = marker.pk / marker_pks.max()
objects_segmented = labels_in.copy()
labels_touching_edge = np.hstack((labels_in[0,:], labels_in[-1,:], labels_in[:,0], labels_in[:,-1]))
labels_touching_edge = np.unique(labels_touching_edge)
is_touching = np.zeros(np.max(labels_in)+1, bool)
is_touching[labels_touching_edge.astype(int)] = True
is_touching = is_touching[labels_in.astype(int)]
labels_in[(~ is_touching) & (objects_segmented == 0)] = 0 ### PRIMARY OBJECTS
# 3. threshold image to be segmented
thresholded_image = threshold_image(img)
# 5. actually do segmentation
labels_out, distance = propagate(img, labels_in, thresholded_image, 0.01)
small_removed_segmented_out = fill_labeled_holes(labels_out)
segmented_out = filter_labels(small_removed_segmented_out, objects_segmented)
return segmented_out
def segment_primary(self, min_size, max_size):
img = self.load().astype(np.float64) / 255.0
# 1. threshold image
binary_image = threshold_image(img)
# 2. fill background holes in foreground objects
def size_fn(size, is_foreground):
return size < max_size
binary_image = fill_labeled_holes(binary_image, size_fn=size_fn)
# 3. perform recognition
labeled_image, object_count = scipy.ndimage.label(binary_image, np.ones((3,3), bool))
labeled_image, object_count, maxima_suppression_size, LoG_threshold, LoG_filter_diameter = separate_neighboring_objects(img, labeled_image, object_count, min_size, max_size)
# 4. fill holes again
labeled_image = fill_labeled_holes(labeled_image)
return labeled_image, object_count
