def compare(function, image, gray_image):
show_original(image)
desc = function_name(function)
for i in range(0, 3):
eroded = function(gray_image.copy(), None, iterations=i + 1)
cv2.imshow(f"{desc} {i + 1} times", eroded)
wait_and_destroy_all_windows()
def compare_average_blur():
"""Fast, but doesn't preserve edges"""
show_original()
for (kx, ky) in kernel_sizes:
blurred_img = cv2.blur(img, (kx, ky))
cv2.imshow("Average ({}, {})".format(kx, ky), blurred_img)
wait_and_destroy_all_windows()
def compare_gaussian_blur():
"""Slightly slower and better at preserving edges"""
show_original()
for (kx, ky) in kernel_sizes:
blurred_img = cv2.GaussianBlur(img, (kx, ky), sigmaX=0, sigmaY=0)
cv2.imshow("Gaussian ({}, {})".format(kx, ky), blurred_img)
wait_and_destroy_all_windows()
def show_otsu_thresholding(image, blurred_image):
(T, thresh_inv) = cv2.threshold(
blurred_image, 0, 255, cv2.THRESH_BINARY_INV | cv2.THRESH_OTSU)
cv2.imshow("Original", image)
cv2.imshow(f"Otsu Thresholded T={T}", thresh_inv)
cv2.imshow("Masked", cv2.bitwise_and(image, image, mask=thresh_inv))
wait_and_destroy_all_windows()
def compare_morphological_op(op_type, elem_type, image, gray_image):
show_original(image)
kernel_sizes = [(3, 3), (5, 5), (7, 7)]
for size in kernel_sizes:
kernel = cv2.getStructuringElement(elem_type, size)
opening = cv2.morphologyEx(gray_image, op_type, kernel)
op_title = morphological_op_name(op_type)
cv2.imshow(f"{op_title}: ({size[0]}, {size[1]})", opening)
wait_and_destroy_all_windows()
def show_contour_masking(image, image_gray):
contours = get_contours(image_gray, cv2.RETR_EXTERNAL)
for idx, contour in enumerate(contours):
mask = np.zeros(image_gray.shape, dtype="uint8")
cv2.drawContours(mask, [contour], -1, (255), -1)
cv2.imshow("Original", image)
cv2.imshow("Mask", mask)
cv2.imshow(f"Masked contours: {idx}",
cv2.bitwise_and(image, image, mask=mask))
wait_and_destroy_all_windows()
def compare_median_blur():
"""
Used to reduce salt-and-pepper style noise as the median statistic is much
less sensitive to outliers than other statistical methods like the mean
"""
show_original()
for k_size in (3, 9, 15):
blurred_img = cv2.medianBlur(img, k_size)
cv2.imshow("Median ({})".format(k_size), blurred_img)
wait_and_destroy_all_windows()
def show_simple_thresholding(image, blurred_image):
(T, thresh_inv) = cv2.threshold(
blurred_image, thresh=200, maxval=255, type=cv2.THRESH_BINARY_INV)
(T, thresh) = cv2.threshold(
blurred_image, thresh=200, maxval=255, type=cv2.THRESH_BINARY)
cv2.imshow("Original", image)
cv2.imshow("Thresholded Inverse", thresh_inv)
cv2.imshow("Thresholded", thresh)
cv2.imshow("Masked", cv2.bitwise_and(image, image, mask=thresh_inv))
wait_and_destroy_all_windows()
def compare_lab(rgb_image):
"""
L* - Lightness, a* - blue-yellow scale, b* red-green scale.
Euclidean distance has perceptual meaning.
"""
lab_image = cv2.cvtColor(rgb_image, cv2.COLOR_BGR2LAB)
cv2.imshow("L*a*b*", lab_image)
for (name, chan) in zip(("L*", "a*", "b*"), cv2.split(lab_image)):
cv2.imshow(name, chan)
wait_and_destroy_all_windows()
def compare_bilateral_blur():
"""Substantially slower, removes detail while preserving edges"""
show_original()
params = [(11, 21, 7), (11, 41, 21), (11, 61, 39)]
for (diameter, sigma_color, sigma_space) in params:
blurred_img = cv2.bilateralFilter(img, diameter, sigma_color,
sigma_space)
title = "Bilateral d={} sc={} ss={}".format(diameter, sigma_color,
sigma_space)
cv2.imshow(title, blurred_img)
wait_and_destroy_all_windows()
def show(function, image, image_gray):
desc = None
if function == cv2.Sobel:
desc = "Sobel"
elif function == cv2.Scharr:
desc = "Scharr"
g_x = function(image_gray, ddepth=cv2.CV_64F, dx=1, dy=0)
g_y = function(image_gray, ddepth=cv2.CV_64F, dx=0, dy=1)
g_x = cv2.convertScaleAbs(g_x)
g_y = cv2.convertScaleAbs(g_y)
image_sobel = cv2.addWeighted(g_x, 0.5, g_y, 0.5, 0)
cv2.imshow("Original", image)
cv2.imshow(f"{desc} X", g_x)
cv2.imshow(f"{desc} Y", g_y)
cv2.imshow(f"{desc} Combined", image_sobel)
wait_and_destroy_all_windows()
def show_all_contours(image, contours):
clone = image.copy()
cv2.drawContours(clone, contours, -1, (0, 255, 0), 2)
cv2.imshow("Original", image)
cv2.imshow("Contours", clone)
wait_and_destroy_all_windows()
def show_external_contours(image, image_gray):
clone = image.copy()
contours = get_contours(image_gray, cv2.RETR_EXTERNAL)
cv2.drawContours(clone, contours, -1, (0, 255, 0), 2)
cv2.imshow("External contours", clone)
wait_and_destroy_all_windows()
def show_each_contour(image, contours):
for (index, contour) in enumerate(contours):
clone = image.copy()
cv2.drawContours(clone, [contour], -1, (0, 255, 0), 2)
cv2.imshow(f"Contour {index}", clone)
wait_and_destroy_all_windows()
def compare_rgb(rgb_image):
cv2.imshow("RGB", rgb_image)
for (name, chan) in zip(("B", "G", "R"), cv2.split(rgb_image)):
cv2.imshow(name, chan)
wait_and_destroy_all_windows()
def compare_hsv(rgb_image):
hsv_image = cv2.cvtColor(rgb_image, cv2.COLOR_BGR2HSV)
cv2.imshow("HSV", hsv_image)
for (name, chan) in zip(("H", "S", "V"), cv2.split(hsv_image)):
cv2.imshow(name, chan)
wait_and_destroy_all_windows()
