Roberts算子、Prewitt算子、Sobel算子、Laplacian算子、Canny算子 代码python

以下是对应算子的简单示例代码：

Roberts算子：

from scipy import ndimage  
import numpy as np

def robert_cross(image):
    roberts_cross_v = np.array( [[-1,0],[0,1]], dtype=np.int32 )
    roberts_cross_h = np.array( [[0,-1],[1,0]], dtype=np.int32 )
    
    vertical = ndimage.convolve( image, roberts_cross_v )
    horizontal = ndimage.convolve( image, roberts_cross_h )
    
    return np.sqrt( np.square(horizontal) + np.square(vertical) )

Prewitt算子：

from scipy import ndimage  
import numpy as np

def prewitt(image):
    vert_prewitt = np.array([[-1,-1,-1],[0,0,0],[1,1,1]])  
    hori_prewitt = np.array([[-1,0,1],[-1,0,1],[-1,0,1]])  
    prewitt_vert = ndimage.convolve(image, vert_prewitt)  
    prewitt_hori = ndimage.convolve(image, hori_prewitt)  
    
    return np.sqrt( np.square(prewitt_vert) + np.square(prewitt_hori) )

Sobel算子：

from scipy import ndimage  
import numpy as np

def sobel(image):
    vert_sobel = np.array([[-1,-2,-1],[0,0,0],[1,2,1]])  
    hori_sobel = np.array([[-1,0,1],[-2,0,2],[-1,0,1]])  
    sobel_vert = ndimage.convolve(image, vert_sobel)  
    sobel_hori = ndimage.convolve(image, hori_sobel)
    
    return np.sqrt( np.square(sobel_vert) + np.square(sobel_hori) )

Laplacian算子：

from scipy import ndimage  
import numpy as np

def laplacian(image):
    laplacian_kernal = np.array([[0, 1, 0], [1, -4, 1], [0, 1, 0]]) 
    filtered_img = ndimage.convolve(image, laplacian_kernal) 
    
    return filtered_img

Canny算子：

from scipy import ndimage  
import numpy as np

def canny(image, low_threshold=0, high_threshold=255):
    img_smooth = ndimage.gaussian_filter(image, sigma=1.4)
    img_x = ndimage.sobel(img_smooth, axis=0)
    img_y = ndimage.sobel(img_smooth, axis=1)
    gradient_magnitude = np.hypot(img_x, img_y)
    gradient_direction = np.arctan2(img_y, img_x)
    gradient_direction = np.rad2deg(gradient_direction)
    gradient_direction[gradient_direction < 0] += 180

    max_magnitude = np.max(gradient_magnitude)
    if high_threshold > max_magnitude:
        high_threshold = max_magnitude
    if low_threshold > high_threshold:
        low_threshold, high_threshold = high_threshold, low_threshold

    row, col = image.shape
    canny_img = np.zeros((row, col), dtype=np.uint8)

    weak = np.int32(50)
    strong = np.int32(255)

    strong_i, strong_j = np.where(gradient_magnitude >= high_threshold)
    zeros_i, zeros_j = np.where(gradient_magnitude < low_threshold)

    weak_i, weak_j = np.where((gradient_magnitude <= high_threshold) & (gradient_magnitude >= low_threshold))

    canny_img[strong_i, strong_j] = strong
    canny_img[weak_i, weak_j] = weak

    for i in range(1, row-1):
        for j in range(1, col-1):
            if canny_img[i,j] == weak:
                if (canny_img[i+1, j-1] == strong) or (canny_img[i+1, j] == strong) or (canny_img[i+1, j+1] == strong) \
                        or (canny_img[i, j-1] == strong) or (canny_img[i, j+1] == strong) \
                        or (canny_img[i-1, j-1] == strong) or (canny_img[i-1, j] == strong) or (canny_img[i-1, j+1] == strong):
                            canny_img[i, j] = strong
                        else:
                            canny_img[i, j] = 0

    return canny_img