def main():
'''
This part is for testing gvf.py with single image.
Input: an original image
Output: Thresholding image and seed image
'''
images = []
temp = cv2.imread(sys.argv[1])
images.append(cv2.cvtColor(temp, cv2.COLOR_BGR2GRAY))
# Binarization
th = athresh(images)
threh = th.applythresh()
# Nuclei center detection
gvf = GVF(images, threh)
dismap = gvf.distancemap()
newimg = gvf.new_image(0.4, dismap) # choose alpha as 0.4.
gradimg = gvf.compute_gvf(newimg)
imgpair = gvf.find_certer(gradimg[0], 0)
cv2.imwrite('imgpair_test.tif', (np.clip(imgpair, 0,
255)).astype(np.uint8))
def main():
use_simple_marker = True
# Binarization
if use_simple_marker:
binarymark, mark = ws_simple(enhance_images)
else:
th = athresh(enhance_images)
threh = th.applythresh()
write_image(threh[0]*255, "threh", 0)
# Nuclei center detection
gvf = GVF(images, threh)
dismap = gvf.distancemap()
newimg = gvf.new_image(10, dismap) # choose alpha as 0.4.
write_image((dismap[0]*10).astype(np.uint8), "dismap", 0)
write_image(newimg[0], "newimg", 0)
gradimg = gvf.compute_gvf(newimg)
if not os.path.exists(temporary_result):
os.makedirs(temporary_result)
os.chdir(temporary_result)
imgpairs = []
# bin_imgpairs = []
# imgpair_raws = []
for i, grad in enumerate(gradimg):
load_img = load_image('imgpair', i)
if load_img == []:
imgpair, bin_imgpair, imgpair_raw = gvf.find_certer(grad, i)
imgpairs.append(imgpair)
# bin_imgpairs.append(bin_imgpair)
# imgpair_raws.append(imgpair_raw)
# write_image(imgpair_raw, 'imgpair_raw', i)
# write_image(bin_imgpair, 'bin_imgpair', i)
write_image(imgpair, 'imgpair', i)
else:
imgpair = load_img
imgpairs.append(imgpair)
os.chdir(os.pardir)
# watershed
ws = WS(newimg, imgpairs)
wsimage, binarymark, mark = ws.watershed_compute()
centroid = []
slope_length = []
# Build Delaunay Triangulation
for i in range(len(images)):
graph = GRAPH(mark, binarymark, i)
tempcentroid, tempslope_length = graph.run(False) # how many centroid points, and slope and length connect to this point
centroid.append(tempcentroid)
slope_length.append(tempslope_length)
print('building Delaunay Triangulation %d/%d '%(i, len(images)))
# Build the Dissimilarity measure vector
vector = []
for i in range(len(images)):
print " feature vector: image ", i
v = FEA()
v.set_centroid(centroid[i])
v.set_spatial(slope_length[i])
v.set_shape(enhance_images[i], mark[i])
v.set_histogram()
v.add_label()
v.add_id(mark[i].max(), i)
gen_vec = v.generate_vector()
vector.append(gen_vec)
print "num of nuclei: ", len(vector[i])
# Feature matching
mask = []
for i in range(len(images)-1):
print " Feature matching: image ", i
m = MAT(i,i+1,[images[i], images[i+1]], vector)
mask.append(m.generate_mask(mark[i], i))
m.find_match(0.3)
mask = m.match_missing(mask, max_frame=2, max_distance=5)
vector[i+1] = m.mitosis_refine()
m.new_id()
vector[i+1] = m.return_vectors()
# write gif image showing the final result
def find_max_id(temp_vector):
max_id = 0
for pv in temp_vector:
for p in pv:
if p.id > max_id:
max_id = p.id
return max_id
# This part is to mark the result in the normolized image and
# write the gif image.
max_id = find_max_id(vector)
max_id = int(max_id)
colors = [np.random.randint(0, 255, size=max_id), \
np.random.randint(0, 255, size=max_id), \
np.random.randint(0, 255, size=max_id)]
font = cv2.FONT_HERSHEY_SIMPLEX
selecy_id = 9
enhance_imgs = []
for i, m in enumerate(mask):
print " write the gif image: image ", i
enhance_imgs.append(cv2.cvtColor(enhance_images[i], cv2.COLOR_GRAY2RGB))
for pv in vector[i]:
center = pv.c
if not pv.l:
color = (colors[0][int(pv.id) - 1], \
colors[1][int(pv.id) - 1], \
colors[2][int(pv.id) - 1],)
else:
color = (colors[0][int(pv.l) - 1], \
colors[1][int(pv.l) - 1], \
colors[2][int(pv.l) - 1],)
if m[int(center[0]), int(center[1])]:
enhance_imgs[i][m == pv.id] = color
cv2.putText(enhance_imgs[i], \
str(int(pv.id)), (int(pv.c[1]), \
int(pv.c[0])),
font, 0.5, \
(255, 255, 255), 1)
os.chdir(temporary_result)
imageio.mimsave('mitosis_final.gif', enhance_imgs, duration=0.6)
cells = set_date(vector)
write_info(cells, "res_track")
render.plot_tracks(vector)
print "finish!"
def main():
# # Binarization
th = athresh(enhance_images)
threh = th.applythresh()
write_image(threh[0] * 255, "threh", 0)
# Nuclei center detection
gvf = GVF(images, threh)
dismap = gvf.distancemap()
newimg = gvf.new_image(4, dismap) # choose alpha as 0.4.
write_image((dismap[0] * 10).astype(np.uint8), "dismap", 0)
write_image(newimg[0], "newimg", 0)
gradimg = gvf.compute_gvf(newimg)
os.chdir("temporary_result")
imgpairs = []
bin_imgpairs = []
imgpair_raws = []
for i, grad in enumerate(gradimg):
imgpair, bin_imgpair, imgpair_raw = gvf.find_certer(grad, i)
imgpairs.append(imgpair)
bin_imgpairs.append(bin_imgpair)
imgpair_raws.append(imgpair_raw)
write_image(imgpair_raw, 'imgpair_raw', i)
write_image(bin_imgpair, 'bin_imgpair', i)
write_image(imgpair, 'imgpair', i)
os.chdir(os.pardir)
# watershed
ws = WS(newimg, imgpair)
wsimage, binarymark, mark = ws.watershed_compute()
centroid = []
slope_length = []
# Build Delaunay Triangulation
for i in range(len(images)):
graph = GRAPH(mark, binarymark, i)
tempcentroid, tempslope_length = graph.run(True)
centroid.append(tempcentroid)
slope_length.append(tempslope_length)
# Build the Dissimilarity measure vector
vector = []
for i in range(len(images)):
print " feature vector: image ", i
v = FEAVECTOR()
v.set_centroid(centroid[i])
v.set_spatial(slope_length[i])
v.set_shape(images[i], markers[i])
v.set_histogram()
v.add_label()
v.add_id(markers[i].max(), i)
vector.append(v.generate_vector())
print "num of nuclei: ", len(vector[i])
# Feature matching
for i in range(len(images) - 1):
print " Feature matching: image ", i
m = SIMPLE_MATCH(i, i + 1, [images[i], images[i + 1]], vector)
mask.append(m.generate_mask(markers[i], i))
m.find_match(0.3)
mask = m.match_missing(mask, max_frame=2, max_distance=20)
vector[i + 1] = m.mitosis_refine()
m.new_id()
vector[i + 1] = m.return_vectors()
# write images if necessary
os.chdir("temporary_result")
for i in range(len(images)):
write_image(imgpair_raw, 'imgpair_raw', i)
write_image(threh[i].astype(np.uint8) * 255, 'threh', i)
write_image(enhance_images[i], 'normalize', i)
write_image(mark[i].astype(np.uint8), 'mark', i)
write_image(binarymark[i].astype(np.uint8), 'binarymark', i)
write_image(tempimg[i].astype(np.uint8), 'tempimg', i)
write_image(dismap[i].astype(np.uint8), 'dismap', i)
os.chdir(os.pardir)