def layer_dict_2_mask(rootDir,
layer_dict,
show=False,
save_dir=None,
show_circle=False):
'''
This function is used to plot the beads on the real tissue image
:param rootDir:
:param layer_dict:
:return:
'''
img_dir = os.path.join(rootDir, "3 - Processed Images",
"7 - Counted Reoriented Stacks Renamed")
def z_key(elem):
return float(elem.split(',')[-1].strip().split('.tif')[0])
tif_list = os.listdir(img_dir)
tif_list.sort(key=z_key)
img = cv2.imread(os.path.join(img_dir, tif_list[0]))
canvas = get_blank_canvas(img)
save_img_list = []
for tif in tif_list:
temp_canvas = canvas.copy()
if show:
show_canvas = temp_canvas.copy()
if z_key(tif) in layer_dict.keys():
layer = layer_dict[z_key(tif)]
for point in layer:
temp_canvas[point[1]][point[0]] = 255
if show_circle:
cv2.circle(temp_canvas,
point,
15, (255, 0, 0),
thickness=5)
if show:
cv2.circle(show_canvas,
point,
15, (255, 0, 0),
thickness=5)
if show:
show_img(show_canvas, tif, True)
temp_canvas = cv2.resize(
temp_canvas,
(int(temp_canvas.shape[1] * 0.5), int(temp_canvas.shape[0] * 0.5)))
save_img_list.append(temp_canvas)
if save_dir is not None:
save_dir = os.path.join(save_dir, "bead")
index = 0
if not os.path.exists(save_dir):
os.mkdir(save_dir)
for i in range(len(save_img_list) - 1, -1, -1):
imsave(os.path.join(save_dir,
str(index) + ".tif"), save_img_list[i])
index += 1
def plot_layer_dict_on_img(rootDir, layer_dict):
'''
This function is used to plot the beads on the real tissue image
:param rootDir:
:param layer_dict:
:return:
'''
img_dir = os.path.join(rootDir, "processed")
def z_key(elem):
return float(elem.split(',')[-1].strip().split('.tif')[0])
tif_list = os.listdir(img_dir)
tif_list.sort(key=z_key)
for tif in tif_list:
if z_key(tif) in layer_dict.keys():
layer = layer_dict[z_key(tif)]
img = cv2.imread(os.path.join(img_dir, tif))
for point in layer:
cv2.circle(img, point, 15, (255, 0, 0), thickness=5)
show_img(img, "bead location", False)
def preprocess_pair(img_frame, atlas_frame, ann_frame, show=False):
'''
Transform the position of the brain in the atlas frame to adapt image frame
:param img_frame:
:param atlas_frame:
:return:
'''
img_frame = cv2.cvtColor(img_frame, cv2.COLOR_BGR2GRAY)
threshold = get_adaptive_threshold(img_frame)
ret, th = cv2.threshold(img_frame, threshold, 255, cv2.THRESH_BINARY)
kernel = np.ones((7, 7), np.uint8)
th = cv2.erode(th, kernel, iterations=2)
_, contours, _ = cv2.findContours(th, cv2.RETR_TREE,
cv2.CHAIN_APPROX_SIMPLE)
tissue_frame = img_frame.copy()
for i in range(len(contours)):
if cv2.contourArea(contours[i]) > 2e5 and cv2.contourArea(
contours[i]) < 1e7:
x_t, y_t, w_t, h_t = cv2.boundingRect(contours[i])
point_t = (int(x_t + 0.5 * w_t), int(y_t + 0.5 * h_t))
if is_in_center(point_t, tissue_frame):
x, y, w, h = cv2.boundingRect(contours[i])
tissue_frame[:, 0:x] = 0
tissue_frame[:, x + w:] = 0
tissue_frame[0:y, :] = 0
tissue_frame[y + h:, :] = 0
mask = np.zeros(tissue_frame.shape).astype(np.uint8)
cv2.drawContours(mask, [contours[i]], -1, 255, -1)
tissue_frame = cv2.bitwise_and(tissue_frame,
tissue_frame,
mask=mask)
if show:
cv2.rectangle(tissue_frame, (x, y), (x + w, y + h),
(255, 255, 0), 5)
show_img(tissue_frame, False)
cur_h = 0
cur_w = 0
(height, length) = atlas_frame.shape
for row in range(height):
if np.asarray(atlas_frame[row, :]).sum() > 0:
cur_h += 1
for col in range(length):
if np.asarray(atlas_frame[:, col]).sum() > 0:
cur_w += 1
try:
w_factor = float(w) / float(cur_w)
h_factor = float(h) / float(cur_h)
except:
tissue_frame = img_frame.copy()
for i in range(len(contours)):
x, y, w, h = cv2.boundingRect(contours[i])
cv2.rectangle(tissue_frame, (x, y), (x + w, y + h), (255, 255, 0),
5)
show_img(tissue_frame, False)
refactored_atlas_center = (int(ATLAS_CERTER_POSITION[2] * w_factor),
int(ATLAS_CERTER_POSITION[1] * h_factor))
atlas_frame = cv2.resize(atlas_frame,
(int(atlas_frame.shape[1] * w_factor),
int(atlas_frame.shape[0] * h_factor)),
interpolation=cv2.INTER_NEAREST)
ann_frame = cv2.resize(ann_frame, (int(
ann_frame.shape[1] * w_factor), int(ann_frame.shape[0] * h_factor)),
interpolation=cv2.INTER_NEAREST)
atlas_frame, atlas_center = get_pure_brain_atlas(atlas_frame,
refactored_atlas_center)
ann_frame, _ = get_pure_brain_atlas(ann_frame, refactored_atlas_center)
canvas_atlas = np.zeros(
(img_frame.shape[0], img_frame.shape[1])).astype(np.uint8)
canvas_ann = np.zeros(
(img_frame.shape[0], img_frame.shape[1])).astype(np.uint16)
atlas_size = atlas_frame.shape
canvas_atlas[0:atlas_size[0], 0:atlas_size[1]] = atlas_frame
ann_size = ann_frame.shape
canvas_ann[0:ann_size[0], 0:ann_size[1]] = ann_frame
canvas_center = (int(canvas_atlas.shape[1] * 0.5),
int(canvas_atlas.shape[0] * 0.5))
shift_col = canvas_center[0] - atlas_center[0]
shift_row = canvas_center[1] - atlas_center[1]
M = np.float32([[1, 0, shift_col], [0, 1, shift_row]])
canvas_atlas = cv2.warpAffine(
canvas_atlas, M, (canvas_atlas.shape[1], canvas_atlas.shape[0]))
canvas_ann = cv2.warpAffine(canvas_ann, M,
(canvas_ann.shape[1], canvas_ann.shape[0]))
return tissue_frame, canvas_atlas, canvas_ann
def preprocess_pair(img_frame, atlas_frame, ann_frame, show=False, strict=0.):
'''
Transform the position of the brain in the atlas frame to adapt image frame
:param img_frame:
:param atlas_frame:
:return:
'''
assert strict <= 1.0 and strict >= 0.
img_frame = cv2.cvtColor(img_frame, cv2.COLOR_BGR2GRAY)
tissue_frame = copy.deepcopy(img_frame)
threshold = get_adaptive_threshold(img_frame)
ret, th = cv2.threshold(img_frame, threshold, 255, cv2.THRESH_BINARY)
kernel = np.ones((7, 7), np.uint8)
th = cv2.erode(th, kernel, iterations=2)
_, contours, _ = cv2.findContours(th, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
contours.sort(key=cv2.contourArea)
# show_img(th, False)
# show_frame = copy.deepcopy(img_frame)
# show_frame = cv2.cvtColor(show_frame, cv2.COLOR_GRAY2RGB)
for i in range(len(contours)):
if cv2.contourArea(contours[i]) > 2e5 and cv2.contourArea(contours[i]) < 5e6:
# x_t, y_t, w_t, h_t = cv2.boundingRect(contours[i])
# point_t = (int(x_t + 0.5 * w_t), int(y_t + 0.5 * h_t))
M = cv2.moments(contours[i])
cX = int(M["m10"] / M["m00"])
cY = int(M["m01"] / M["m00"])
# show_frame = cv2.drawContours(show_frame, [contours[i]], 0, (0,255,0), 3)
# show_frame = cv2.circle(show_frame, (cX, cY), 3, (0, 0, 255), 3)
if is_in_center((cX, cY), tissue_frame):
x, y, w, h = cv2.boundingRect(contours[i])
# cv2.rectangle(show_frame, (x - 75, y - 75), (x + w + 150, y + h + 150), (255, 0, 0), thickness=3, lineType=8)
# show_img(th, False)
# convex_contours = cv2.convexHull(contours[i])
# mask = np.zeros(tissue_frame.shape).astype(np.uint8)
# cv2.drawContours(mask, [convex_contours], -1, 255, -1)
# tissue_frame = cv2.bitwise_and(tissue_frame, tissue_frame, mask=mask)
color_frame = cv2.cvtColor(tissue_frame, cv2.COLOR_GRAY2RGB)
# rect = (x, y, w + x, h + y)
rect = (x - 75, y - 75, w + 150, h + 150)
mask = np.zeros(tissue_frame.shape,np.uint8)
bgdModel = np.zeros((1,65),np.float64)
fgdModel = np.zeros((1,65),np.float64)
cv2.grabCut(color_frame, mask, rect, bgdModel, fgdModel, 5, cv2.GC_INIT_WITH_RECT)
# mask2 = np.where((mask==2)|(mask==0),0,1).astype('uint8')
mask2 = np.where((mask==2)|(mask==0), 1.0 - strict, 1)
tissue_frame[:, 0: x - 75] = 0
tissue_frame[:, x + w + 75:] = 0
tissue_frame[0:y - 75, :] = 0
tissue_frame[y + h + 75:, :] = 0
tissue_frame = tissue_frame * mask2[:,:]
tissue_frame = np.asarray(tissue_frame).astype(np.uint8)
if show:
print("normal")
cv2.rectangle(tissue_frame, (x, y), (x + w, y + h), (255, 255, 0), 5)
show_img(tissue_frame, False)
# show_img(tissue_frame, False)
atlas_frame = np.asarray(atlas_frame, dtype=np.float32)
atlas_frame = ((atlas_frame - atlas_frame.min()) / (atlas_frame.max() - atlas_frame.min())) * 255
atlas_frame = atlas_frame.astype(np.uint8)
ann_frame, (row1, row2, col1, col2) = get_pure_brain_atlas(ann_frame, threshold=0)
atlas_frame = atlas_frame[row1:row2, col1:col2]
ori_ann_center = (ATLAS_CERTER_POSITION[2] - row1, ATLAS_CERTER_POSITION[1] - col1)
cur_w = col2 - col1
cur_h = row2 - row1
print(ori_ann_center)
try:
w_factor = float(w) / float(cur_w)
h_factor = float(h) / float(cur_h)
except:
tissue_frame = img_frame.copy()
for i in range(len(contours)):
x, y, w, h = cv2.boundingRect(contours[i])
cv2.rectangle(tissue_frame, (x, y), (x + w, y + h), (255, 255, 0), 5)
print("exception")
show_img(tissue_frame, False)
# print(w_factor, h_factor)
refactored_atlas_center = (int(ori_ann_center[1] * w_factor), int(ori_ann_center[0] * h_factor))
atlas_frame = cv2.resize(atlas_frame, (int(atlas_frame.shape[1] * w_factor), int(atlas_frame.shape[0] * h_factor)), interpolation=cv2.INTER_NEAREST)
ann_frame = cv2.resize(ann_frame, (int(ann_frame.shape[1] * w_factor), int(ann_frame.shape[0] * h_factor)), interpolation=cv2.INTER_NEAREST)
canvas_atlas = np.zeros((img_frame.shape[0], img_frame.shape[1])).astype(np.uint8)
canvas_ann = np.zeros((img_frame.shape[0], img_frame.shape[1])).astype(np.uint16)
atlas_size = atlas_frame.shape
canvas_atlas[0:atlas_size[0], 0:atlas_size[1]] = atlas_frame
ann_size = ann_frame.shape
canvas_ann[0:ann_size[0], 0:ann_size[1]] = ann_frame
# show_img(canvas_atlas, False)
canvas_center = (int(canvas_atlas.shape[1] * 0.5), int(canvas_atlas.shape[0] * 0.5))
shift_col = canvas_center[0] - refactored_atlas_center[0]
shift_row = canvas_center[1] - refactored_atlas_center[1]
M = np.float32([[1, 0, shift_col], [0, 1, shift_row]])
canvas_atlas = cv2.warpAffine(canvas_atlas, M, (canvas_atlas.shape[1], canvas_atlas.shape[0]))
canvas_ann = cv2.warpAffine(canvas_ann, M, (canvas_ann.shape[1], canvas_ann.shape[0]))
# show_img(canvas_atlas, False)
return tissue_frame, canvas_atlas, canvas_ann