def find_square_ragion(gray_image):
# for some big size pictures which need to blur,
# but cannot using ksize=(5,5), since some picture will get wrong number value.
blured_image = cv2.GaussianBlur(gray_image, ksize=(3,3), sigmaX=0)
# Display.image(blured_image)
threshed_image = cv2.adaptiveThreshold(blured_image,WHITE,
cv2.ADAPTIVE_THRESH_MEAN_C,cv2.THRESH_BINARY_INV, blockSize=7, C=2)
# flag_test()
# Display.image(threshed_image)
'''
It's very depend on the threshed_image.
'''
max_contour = find_max_contour(threshed_image)
# Display.contours(gray_image, [max_contour])
# max_contour.ppl()
# 'test'.pl()
# from picture_sudoku.helpers.common import Resource
# numpy.save(Resource.get_path("for_issues/max_contour.npy"), max_contour)
# square_contour = convert_to_square(max_contour)
# this is using the border line.
square_contour = extract_square_from_contour(max_contour)
# square_contour.ppl()
if square_contour == None:
raise SudokuError("Cannot find sudoku square!")
# flag_test()
# Display.contours(gray_image, [square_contour])
larged_contour = Quadrilateral.enlarge(square_contour, 0.007)
larged_contour = Contour.check_beyond_borders(larged_contour, gray_image.shape)
# Display.contours(gray_image, [larged_contour])
# larged_contour.ppl()
square_ragion = Contour.get_rect_ragion(larged_contour, gray_image)
# Display.image(square_ragion)
return square_ragion
