Convolve a N-D kernel with a NxN matrix, such as an image

convolveMat.py

import numpy as np
import matplotlib.image as img
import matplotlib.pyplot as plt

image = img.imread('school.png') # Grab the image

def toGreyscale(image):
    # convert the RGB to Greyscale
    return np.dot(image[...,:3], [0.114, 0.299, 0.587])

image = toGreyscale(image) 

if 0:
    plt.imshow(image)
    plt.show()

kernel = np.array([[-1,-1,1],[-1,1,-1],[1,-1,-1]]) # kernel
img = image

# Apply the kernel
# Inspired from: https://stackoverflow.com/questions/43373521/how-to-do-convolution-matrix-operation-in-numpy
x,y = kernel.shape
x_dif, y_dif = np.subtract(img.shape,kernel.shape)
lst = []
for x_i in range(1+x_dif):
    for y_i in range(1+y_dif):
        lst.append(np.sum(np.multiply(img[x_i:x+x_i,y_i:y+y_i],kernel)))
        
result_shape = np.add((x_dif,y_dif),(1,1))
result = np.reshape(lst, result_shape)

if 1:
    plt.imshow(result)
    plt.show()