def get_derivatives(self, I_x, I_y):
"""
Get the edges using derivative masks.
:I_x: x-component of the input image.
:I_y: y-component of the input image.
"""
if (self.dimension == 0):
# Define the 1-D derivative mask.
G = np.array([-1, 0, 1])
# Compute the derivatives in both components.
I_prime_x = convolve1d(I_x, G, axis=0)
I_prime_y = convolve1d(I_y, G, axis=1)
elif (self.dimension == 1):
# Define the 2-D derivative masks.
G_x = np.array([[-1, 0, 1], [-1, 0, 1], [-1, 0, 1]])
G_y = np.array([[-1, -1, -1], [0, 0, 0], [1, 1, 1]])
# Calculate magnitude of the input.
I = np.hypot(I_x, I_y).astype(int)
# Compute the derivatives in both components.
I_prime_x = ndimage.filters.convolve(I, G_x)
I_prime_y = ndimage.filters.convolve(I, G_y)
else:
print("Error: Invalid dimension.")
return
# Calculate the magnitude.
I_prime = np.hypot(I_prime_x, I_prime_y).astype(int)
I_prime = I_prime / I_prime.max() * 255
# Calculate the direction.
I_theta = np.arctan2(I_prime_y, I_prime_x).astype(int)
return (I_prime, I_prime_x, I_prime_y, I_theta)
def gaussian_smoothing(self, I):
"""
Perform 1-D Gaussian smoothing.
:I: Input image.
"""
# Define the range of the Gaussian distribution.
r = np.linspace(-(self.kernel_size // 2), self.kernel_size // 2,
self.kernel_size)
# Calculate the Gaussian mask.
G = [(np.exp((-x**2.0) / (2.0 * (self.sigma**2.0))) *
(1 / (np.sqrt(2.0 * np.pi * (self.sigma**2.0))))) for x in r]
# Convolve for both axes.
I_smooth_x = convolve1d(I, G, axis=0)
I_smooth_y = convolve1d(I, G, axis=1)
# Combine both components.
I_smooth = np.hypot(I_smooth_x, I_smooth_y).astype(int)
return (I_smooth, I_smooth_x, I_smooth_y)
def blur_image_kernel1D(image, kernel):
blurred = utils.convolve1d(image, kernel, axis=0)
blurred = utils.convolve1d(blurred, kernel, axis=1)
return blurred
def blur_image(image, radius):
k = create_1d_gaussian_kernel(radius)
blurred = utils.convolve1d(image, k, axis=0)
blurred = utils.convolve1d(blurred, k, axis=1)
return blurred
