def temp_image_folder_resize(input_f, output_f, pixel=(2160, 2160), box=None):
# pixel = (w, h)
# box = (x0,y0,x1,y1)
if not os.path.exists(input_f):
print('!ERROR! The input_f does not existed!')
return False
if not os.path.exists(output_f):
os.makedirs(output_f)
path_list = os.listdir(input_f)
for img_name in path_list:
img = os.path.join(input_f, img_name)
img = image_to_gray(img)
if box is not None:
img = img[box[1]:box[3], box[0]:box[2]]
h, w = img.shape[:2]
if h != pixel[1] or w != pixel[0]:
img = cv2.resize(img, pixel, interpolation=cv2.INTER_NEAREST)
to_file = os.path.join(output_f, img_name)
cv2.imwrite(to_file, img)
return True
def painting_well_segmentation(labels, in_folder, out_folder, files_sort_function, safe_time=5):
if not os.path.exists(in_folder):
print('!ERROR! The in_folder does not existed!')
return False
# time.sleep(safe_time)
if not os.path.exists(out_folder):
# shutil.rmtree(out_folder)
time.sleep(safe_time)
os.makedirs(out_folder)
time.sleep(safe_time)
img_type = ['jpg', 'png', 'tif']
offset = 0
in_files = os.listdir(in_folder)
in_files = files_sort_function(in_files)
for f in in_files:
if f.split('.')[1] in img_type:
in_img_f = os.path.join(in_folder, f)
in_img = image_to_gray(in_img_f)
pixel = in_img.shape[0] * in_img.shape[1]
this_labels = labels[offset:offset + pixel]
offset = offset + pixel
out_file = os.path.join(out_folder, f)
plt.imsave(out_file, this_labels.reshape(in_img.shape))
return True
def get_folder_negative_film_V3(img1, img2, sub_folder, result_path, sort_function, save=True, show=False):
# the copy of V2
# img is the Brightest value of img1 and the Darkest value of img2
# img sub every image in sub_folder, and save to result_path
if not os.path.exists(sub_folder):
print('!ERROR! The sub_folder does not existed!')
return False
if not os.path.exists(result_path):
os.makedirs(result_path)
img1 = image_to_gray(img1)
img2 = image_to_gray(img2)
h1, w1 = img1.shape[:2]
h2, w2 = img2.shape[:2]
if h1 != h2 or w1 != w2:
img2 = cv2.resize(img2, (w1, h1), interpolation=cv2.INTER_NEAREST)
img = img2
x = (img1 >= 128) & (img2 >= 128) & (img1 >= img2)
img[x] = img1[x]
path_list = os.listdir(sub_folder)
sort_function(path_list)
for this_img_name in path_list:
this_img = os.path.join(sub_folder, this_img_name)
this_img = image_to_gray(this_img)
this_h, this_w = this_img.shape[:2]
if h1 != this_h or w1 != this_w:
this_img = cv2.resize(this_img, (w1, h1), interpolation=cv2.INTER_NEAREST)
result = img - this_img
result[result < 0] = 0
# result = abs(result)
if save:
# plt.savefig(os.path.join(result_path, name))
to_file = os.path.join(result_path, this_img_name)
cv2.imwrite(to_file, result)
if show:
cv2.imshow('image_sub', result)
cv2.waitKey(0)
cv2.destroyAllWindows()
return True
def get_folder_negative_film_V1(img1, img2, sub_folder, result_path, sort_function, save=True, show=False):
# img1 sub every image in sub_folder, and save to result_path
# if not os.path.exists(main_path):
# print('!ERROR! The main_path does not existed!')
# return False
# sub_folder = os.path.join(main_path, sub_folder)
if not os.path.exists(sub_folder):
print('!ERROR! The sub_folder does not existed!')
return False
# result_path = os.path.join(main_path, result_path)
if not os.path.exists(result_path):
os.makedirs(result_path)
img1 = image_to_gray(img1)
h1, w1 = img1.shape[:2]
path_list = os.listdir(sub_folder)
sort_function(path_list)
for img2_name in path_list:
img2 = os.path.join(sub_folder, img2_name)
print('>>> get_folder_negative_film_V1(', img2, ')')
img2 = image_to_gray(img2)
h2, w2 = img2.shape[:2]
if h1 != h2 or w1 != w2:
img2 = cv2.resize(img2, (w1, h1), interpolation=cv2.INTER_NEAREST)
result = img1 - img2
result[result < 0] = 0
# result = abs(result)
if save:
# plt.savefig(os.path.join(result_path, name))
to_file = os.path.join(result_path, img2_name)
cv2.imwrite(to_file, result)
if show:
cv2.imshow('image_sub', result)
cv2.waitKey(0)
cv2.destroyAllWindows()
return True
def get_well_difference(input_f, output_f, sort_function):
# calculate adjacent pictures' difference of one folder
# notice: it must giving the sort function
# if not os.path.exists(main_path):
# print('!ERROR! The main_path does not existed!')
# return False
# folder = os.path.join(main_path, folder)
if not os.path.exists(input_f):
print('!ERROR! The input_f does not existed!')
return False
# result_path = os.path.join(main_path, result_path)
if not os.path.exists(output_f):
os.makedirs(output_f)
path_list = os.listdir(input_f)
path_list = sort_function(path_list)
for i in range(len(path_list) - 1):
img1 = os.path.join(input_f, path_list[i])
img1 = image_to_gray(img1)
img2 = os.path.join(input_f, path_list[i + 1])
img2 = image_to_gray(img2)
h1, w1 = img1.shape[:2]
h2, w2 = img2.shape[:2]
if h1 != h2 or w1 != w2:
img2 = cv2.resize(img2, (w1, h1), interpolation=cv2.INTER_NEAREST)
result = img1 - img2
result[result < 0] = 0
# result = abs(result)
this_range = np.max(result) - np.min(result)
result = (result - np.min(result)) / this_range
result = result * 255
to_file = os.path.join(output_f, path_list[i])
cv2.imwrite(to_file, result)
return True
def get_image_image_sub(main_path, img1, img2, result_path='image_sub', name='sub.png', save=True, show=False):
if not os.path.exists(main_path):
print('!ERROR! The main_path does not existed!')
return False
result_path = os.path.join(main_path, result_path)
if not os.path.exists(result_path):
os.makedirs(result_path)
img1 = image_to_gray(img1)
img2 = image_to_gray(img2)
result = img1 - img2
result[result < 0] = 0
# result = abs(result)
if save:
# plt.savefig(os.path.join(result_path, name))
to_file = os.path.join(result_path, name)
cv2.imwrite(to_file, result)
if show:
cv2.imshow('image_sub', result)
cv2.waitKey(0)
cv2.destroyAllWindows()
return True
def get_image_hist(main_path, well_image, result_path='Hist', name='Image_Hist.png', save=True, show=False):
if not os.path.exists(main_path):
print('!ERROR! The main_path does not existed!')
return False
result_path = os.path.join(main_path, result_path)
if not os.path.exists(result_path):
os.makedirs(result_path)
img = image_to_gray(well_image)
h, w = img.shape[:2]
pixelSequence = img.reshape([h * w, ])
histogram, bins, patch = plt.hist(pixelSequence, 256, facecolor='black', histtype='bar')
plt.xlabel("gray label")
plt.ylabel("number of pixels")
plt.axis([0, 255, 0, np.max(histogram)])
if save:
plt.savefig(os.path.join(result_path, name))
if show:
plt.show()
return True
def get_image_auto_binary(main_path, well_image, result_path='auto_binary', inv=False, show=False, blockSize=None,
C=16):
# using cv2.adaptiveThreshold to enhance image(auto intensity threshold)
# once one imgae
# NOTICE: blockSize and C
# blockSize is the MEAN or GAUSSIAN kernal size
# C is a constant: Distinguish distance
# if inv is True then do Negative Film
if not os.path.exists(main_path):
print('!ERROR! The main_path does not existed!')
return False
if not os.path.exists(well_image):
print('!ERROR! The well_image does not existed!')
return False
result_path = os.path.join(main_path, result_path)
if not os.path.exists(result_path):
os.makedirs(result_path)
t_path_list = os.path.split(well_image) # [r'D:\pro\CD22\SSS_100%\S1', '2018-11-28~IPS_CD13~T1.jpg']
t1_path_list = os.path.split(t_path_list[0]) # [r'D:\pro\CD22\SSS_100%', 'S1']
t2_path_list = os.path.split(t1_path_list[0]) # [r'D:\pro\CD22', 'SSS_100%']
SSS_folder = t2_path_list[1] # 'SSS_100%'
S_index = t1_path_list[1] # 'S1'
S = int(S_index.split('S')[1]) # 1
img_name = t_path_list[1] # '2018-11-28~IPS_CD13~T1.png'
name_index = img_name[:-4] # '2018-11-28~IPS_CD13~T1'
# T = int(name_index.split('~T')[1]) # 1
# if S <= 2:
# return True
print('>>>', S_index, '---', name_index, ' Image Auto Binary :::')
to_file = os.path.join(result_path, SSS_folder, S_index, img_name)
if not os.path.exists(os.path.join(result_path, SSS_folder)):
os.makedirs(os.path.join(result_path, SSS_folder))
if not os.path.exists(os.path.join(result_path, SSS_folder, S_index)):
os.makedirs(os.path.join(result_path, SSS_folder, S_index))
img = image_to_gray(well_image)
h, w = img.shape[:2]
if blockSize is None:
blockSize = int(np.min([h, w]) / 10)
if blockSize % 2 != 1:
blockSize = blockSize - 1
if C is None:
C = int(abs(np.mean(img) - np.median(img)))
if C == 0:
C = 16
print('blockSize is: ', blockSize, 'C is: ', C)
if inv:
result_th = cv2.adaptiveThreshold(img, 255, cv2.ADAPTIVE_THRESH_MEAN_C, cv2.THRESH_BINARY_INV, blockSize, C)
# result_th = cv2.adaptiveThreshold(img, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY_INV, blockSize, C)
else:
result_th = cv2.adaptiveThreshold(img, 255, cv2.ADAPTIVE_THRESH_MEAN_C, cv2.THRESH_BINARY, blockSize, C)
# result_th = cv2.adaptiveThreshold(img, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, blockSize, C)
cv2.imwrite(to_file, result_th)
if show:
cv2.imshow('adaptiveThreshold_image', result_th)
cv2.waitKey(0)
cv2.destroyAllWindows()
return True
def return_mask(main_path, sigle_image, kernel=(69, 69), pix=(23, 23), getF=getMean, getC=get_Bright_Dark,
n_clusters=3, result_path='pix_features', save=True, save_name=None, show=False):
# this program is design for core analysis; one time one well one time point one image
# for block feature extraction and cluster
# kernel is feature extraction unit, the window
# pix is compressed image block, the Super pixel
# kernel >= pix
# now it is doing feature extraction
# input sigle_image has well edge, is 5*5
# sigle_image is a image path
# kernel or pix is (w,h)(x,y)
if save:
result_path = os.path.join(main_path, result_path)
if os.path.exists(result_path):
# shutil.rmtree(output_folder)
pass
else:
os.makedirs(result_path) # make the output folder
if type(sigle_image) is str:
sigle_image = os.path.join(main_path, sigle_image)
if save_name is None:
save_name = os.path.split(sigle_image)[1]
img = image_to_gray(sigle_image)
if show:
print(img.shape)
print(img[int(img.shape[0] / 2):int(img.shape[0] / 2 + img.shape[0] / 10),
int(img.shape[1] / 2):int(img.shape[1] / 2 + img.shape[1] / 10)])
cv2.imshow('image_to_gray()', img)
cv2.waitKey()
cv2.destroyAllWindows()
if img is None:
return None
pix_h = img.shape[0] # y:height,row
pix_w = img.shape[1] # x:width,col
compress_pix_w = int((pix_w - kernel[0] + pix[0]) / pix[0]) # x:width,col
compress_pix_h = int((pix_h - kernel[1] + pix[1]) / pix[1]) # y:height,row
offset_x = int((kernel[0] - pix[0]) / 2) # x:width,col
offset_y = int((kernel[1] - pix[1]) / 2) # y:height,row
# used_pix_w = compress_pix_w * pix[0] - pix[0] + kernel[0]
# used_pix_h = compress_pix_h * pix[1] - pix[1] + kernel[1]
# used_img = img[:used_pix_h, :used_pix_w, :]
if show:
print('compress_pix_w = ', compress_pix_w)
print('compress_pix_h = ', compress_pix_h)
pix_features_list = [] # is a col of
for row in range(compress_pix_h):
for col in range(compress_pix_w):
stamp = img[row * pix[1]:row * pix[1] + kernel[1], col * pix[0]:col * pix[0] + kernel[0]]
# if show:
# cv2.imshow('steps(' + str(row) + ',' + str(col) + ')', stamp)
# cv2.waitKey()
# cv2.destroyAllWindows()
pix_features_list.append(getF(stamp))
pix_features_list = np.array(pix_features_list)
pix_features_list = np.nan_to_num(pix_features_list)
# print(pix_features_list)
# print(pix_features_list.shape)
label = getC(pix_features_list, n_clusters=n_clusters)
# label = pix_features_list[:,0]
label_map = label.reshape((compress_pix_h, compress_pix_w))
# label_map = np.array(255 / (label_map + 1), dtype=np.uint8)
# label_map_colored = np.zeros((compress_pix_h, compress_pix_w, 3), dtype=np.uint8)
# color_key = return_color()
# for i in range(compress_pix_h):
# for j in range(compress_pix_w):
# label_map_colored[i, j] = color_key[int(label_map[i, j])]
# cv2.resize(img_cpy, (compress_pix_w*pix[0], compress_pix_h*pix[1]), interpolation=cv2.INTER_NEAREST)
label_map_big = cv2.resize(label_map, (0, 0), fx=pix[0], fy=pix[1], interpolation=cv2.INTER_NEAREST)
print(label_map_big.shape)
label_result = np.ones((pix_h, pix_w), dtype=np.uint8) * 128
print(label_result.shape)
label_result[offset_y:offset_y + compress_pix_h * pix[1],
offset_x:offset_x + compress_pix_w * pix[0]] = label_map_big
if save:
if save_name is None:
save_name = 'result.png'
result_file = os.path.join(result_path, save_name)
cv2.imwrite(result_file, label_result)
if show:
cv2.imshow('get_img_pix_features_list()', label_result)
cv2.waitKey()
cv2.destroyAllWindows()
return False