def cal_assignment_fea(assignment, desc):
cnt1, cnt2 = assignment[0], assignment[1]
arr1 = cv2_transform.cv_cnt_to_np_arr(cnt1)
arr2 = cv2_transform.cv_cnt_to_np_arr(cnt2)
mask1 = img.build_cnt_mask(arr1)
mask2 = img.build_cnt_mask(arr2)
fea1 = desc.describe(mask1)
fea2 = desc.describe(mask2)
return fea1, fea2
def test_smooth_cnt():
cnt1 = np.array([[300, 400, 450, 400, 300, 200, 0, 50, 100, 200],
[100, 100, 200, 300, 400, 500, 500, 400, 300, 200]])
# img1 = build_cnt_mask(cnt1)
smooth_cnt1 = smooth_cnt(cnt1, sigma=10)
smooth_img1 = build_cnt_mask(smooth_cnt1)
import matplotlib.pyplot as plt
plt.imshow(smooth_img1)
def test_build_cnt_mask():
np_arr = np.array([[10, 40, 50, 40, 20], [10, 20, 30, 40, 30]])
# np_arr = np.array([[10, 25, 30, 10], [10, 15, 30, 30]])
mask = build_cnt_mask(np_arr)
# mask = build_cnt_mask(np_arr, mask_size=(100, 100))
# import matplotlib.pyplot as plt
# plt.imshow(mask)
# plt.show()
# import pdb; pdb.set_trace()
_ = np.count_nonzero(mask == 255)
# calculate the contour area
cv_cnt = np_arr_to_cv_cnt(np_arr)
_ = get_cnt_area(cv_cnt)
def get_chromosome_shape(chromosome_path, thresh=28):
chromosome_name = os.path.splitext(os.path.basename(chromosome_path))[0]
karyotype_name = os.path.basename(os.path.dirname(chromosome_path))
ch_img = io.imread(chromosome_path)
ch_bin = ch_img > thresh
ch_bin = binary_fill_holes(ch_bin)
ch_mask = img_as_ubyte(ch_bin)
try:
_, cnts, _ = cv2.findContours(ch_mask,
mode=cv2.RETR_EXTERNAL,
method=cv2.CHAIN_APPROX_NONE)
except:
import pdb
pdb.set_trace()
refine_cnts = list(filter(lambda x: cv2.contourArea(x) > 48, cnts))
refine_cnts.sort(key=(lambda cnt: np.mean(cnt[:, 0, 0])))
if not (len(refine_cnts) <= 2 and len(refine_cnts) >= 0):
print("In {}, the number of contours in {} is {}".format(
karyotype_name, chromosome_name, len(refine_cnts)))
return None
chromosome_desc = {}
desc = ZernikeMoments(21)
for cur_cnt in refine_cnts:
cur_arr = cv2_transform.cv_cnt_to_np_arr(cur_cnt)
shape_mask = img.build_cnt_mask(cur_arr)
# cur_arr_min_h = np.min(cur_arr[0])
# cur_arr_min_w = np.min(cur_arr[1])
# mask_h, mask_w = shape_mask.shape
# sub_img = ch_img[cur_arr_min_h:cur_arr_min_h+mask_h,
# cur_arr_min_w:cur_arr_min_w+mask_w]
# shift_cnt = cur_cnt.copy()
# shift_cnt[:, 0, 0] -= cur_arr_min_w
# shift_cnt[:, 0, 1] -= cur_arr_min_h
# sub_img3 = np.stack((sub_img, sub_img, sub_img), axis=-1)
# cv2.drawContours(sub_img3, [shift_cnt], 0, [127, 0, 0], 2)
# io.imsave('boundary.png', sub_img3)
cur_fea = desc.describe(shape_mask)
shape_name = chromosome_name + '_' + str(uuid.uuid4())[:8]
chromosome_desc[shape_name] = list(cur_fea)
return chromosome_desc