def marr_hildreth_edge_detector(image, sigma):
image_array = np.asarray(image)
# Smooth the image by Gaussian filter
smoothed_image_array = gc.gaussconvolve2d(image_array, sigma)
# Apply Laplacian to smoothed image
laplacian_of_img = laplacian(smoothed_image_array)
# Find zero crossings
result = find_zero_crossings(laplacian_of_img)
return result
def canny_edge_detector(image, sigma):
# Convert to numpy array
image_array = np.asarray(image)
result_array = np.zeros((image_array.shape[0], image_array.shape[1]), dtype=np.int)
# Smooth image by Gaussian filter
smoothed_image_array = gc.gaussconvolve2d(image_array, sigma)
# Compute derivative of filtered image
dx = cid.xderivative(smoothed_image_array.copy())
dy = cid.yderivative(smoothed_image_array.copy())
# Find magnitude and orientation of gradient
row = []
gradient_array = []
# Match up the rows
for ix, iy in zip(dx, dy):
# For each row, match up columns
for ixx, iyy in zip(ix, iy):
gradient_magnitude = math.sqrt((math.pow(ixx,2) + math.pow(iyy,2)))
gradient_direction = np.arctan2(iyy, ixx)
row.append(np.asarray([gradient_magnitude, gradient_direction]))
gradient_array.append(row)
row = []
gradient_array = np.asarray(gradient_array)
# Apply non-maximum suppression
for row in range(0, len(gradient_array)):
for col in range(0, len(gradient_array[0])):
pixel = gradient_array[row][col]
if(is_maximum(row, col, pixel, gradient_array)):
result_array[row][col] = 120
# Apply hysteresis thresholding
result_array = apply_hysteresis_thresholding(gradient_array, result_array)
return result_array
