def segment_image(plate_image: ndarray,
plate_mask: ndarray = None,
plate_noise_mask: ndarray = None,
area_min: float = 1) -> ndarray:
"""
Attempts to separate and label all colonies on a plate
:param plate_image: an image containing colonies
:param plate_mask: a boolean image mask to remove from the original image
:param plate_noise_mask: a black and white image as a numpy array
:param area_min: the minimum area for a colony, in pixels
:returns: a segmented and labelled image as a numpy array
"""
from numpy import unique, isin
from skimage.measure import regionprops, label
from skimage.morphology import remove_small_objects, binary_erosion
from skimage.segmentation import clear_border
plate_image = imaging.remove_background_mask(plate_image, smoothing=0.5)
if plate_mask is not None:
# Remove mask from image
plate_image = plate_image & plate_mask
# Remove objects touching the mask border
plate_image = clear_border(plate_image,
bgval=0,
mask=binary_erosion(plate_mask))
else:
# Remove objects touching the image border
plate_image = clear_border(plate_image, buffer_size=2, bgval=0)
plate_image = label(plate_image, connectivity=2)
# Remove background noise
if len(unique(plate_image)) > 1:
plate_image = remove_small_objects(plate_image, min_size=area_min)
# Remove colonies that have grown on top of image artefacts or static objects
if plate_noise_mask is not None:
plate_noise_image = imaging.remove_background_mask(plate_noise_mask,
smoothing=0.5)
if len(unique(plate_noise_mask)) > 1:
noise_mask = remove_small_objects(plate_noise_image,
min_size=area_min)
# Remove all objects where there is an existing static object
exclusion = unique(plate_image[noise_mask])
exclusion_mask = isin(plate_image, exclusion[exclusion > 0])
plate_image[exclusion_mask] = 0
return plate_image
def segment_image(plate_image, plate_mask, plate_noise_mask, area_min = 5):
"""
Finds all colonies on a plate and returns an array of co-ordinates
If a co-ordinate is occupied by a colony, it contains that colonies labelled number
:param plate_image: a black and white image as a numpy array
:param mask: a black and white image as a numpy array
:param plate_noise_mask: a black and white image as a numpy array
:returns: a segmented and labelled image as a numpy array
"""
from math import pi
from scipy.ndimage.morphology import binary_fill_holes
from skimage.morphology import remove_small_objects
from skimage.measure import regionprops, label
plate_image = imaging.remove_background_mask(plate_image, plate_mask)
plate_noise_mask = imaging.remove_background_mask(plate_noise_mask, plate_mask)
# Subtract an image of the first (i.e. empty) plate to remove static noise
plate_image[plate_noise_mask] = 0
# Fill any small gaps
plate_image = binary_fill_holes(plate_image)
# Remove background noise
plate_image = remove_small_objects(plate_image, min_size = area_min)
colonies = label(plate_image)
# Remove colonies that are on the edge of the plate
# versions <0.16 do not allow for a mask
# colonies = clear_border(pl_th, buffer_size = 1, mask = plate_mask)
# Exclude objects that are too eccentric
rps = regionprops(colonies)
for rp in rps:
# Eccentricity of zero is a perfect circle
# Circularity of 1 is a perfect circle
circularity = (4 * pi * rp.area) / (rp.perimeter * rp.perimeter)
if rp.eccentricity > 0.6 or circularity < 0.85:
colonies[colonies == rp.label] = 0
return colonies
def test_remove_background(self, image_segmented_local, image_mask):
image_ref = np.array([[0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 0, 0, 0, 0],
[0, 0, 0, 1, 0, 1, 0, 0, 0],
[0, 0, 0, 1, 1, 1, 1, 0, 0],
[0, 1, 1, 0, 1, 1, 1, 1, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0]])
result = remove_background_mask(image_segmented_local, image_mask)
assert result.shape == image_segmented_local.shape
assert (result == image_ref).all()
def test_mask_empty(self, image_segmented_local):
with pytest.raises(ValueError):
remove_background_mask(image_segmented_local, np.array([]))
def test_image_empty(self, image_mask):
with pytest.raises(ValueError):
remove_background_mask(np.array([]), image_mask)
def test_image_blank(self):
image_blank = np.zeros((3, 3))
result = remove_background_mask(image_blank, image_blank > 0)
assert (result == image_blank).all()
def test_size_mismatch(self):
with pytest.raises(ValueError):
remove_background_mask(np.ones((5, 5), dtype=np.uint8),
np.ones((3, 5), dtype=np.uint8))
