B Draft

import numpy as np

def pad_image(image: np.ndarray, pad_size: int) -> np.ndarray:
    if pad_size < 1:
        raise ValueError
    res = np.array([])
    m = image.shape[0]
    n = image.shape[1]
    if image.ndim == 2:
        res = np.zeros(shape = (m + 2 * pad_size, n + 2 * pad_size))
    elif image.ndim == 3:
        k = image.shape[2]
        res = np.zeros(shape=(m + 2 * pad_size, n + 2 * pad_size, k))
    res[pad_size: pad_size + m, pad_size: pad_size + n] = image
    return res

def blur_image(
    image: np.ndarray,
    kernel_size: int,
) -> np.ndarray:
    if kernel_size < 1 or kernel_size % 2 == 0:
        raise ValueError
    pad_size = kernel_size // 2
    padded_image = pad_image(image, pad_size)
    N = image.shape[0]
    M = image.shape[1]
    res = np.zeros(shape = image.shape)
    print("image")
    print(image)
    print("padded_image")
    print(padded_image)
    if image.ndim == 2:
        for i in range(N):
            for j in range(M):
                sliding_window = padded_image[i : i + kernel_size, j : j + kernel_size]
                res[i, j] = np.mean(sliding_window)
    else:
        K = image.shape[2]
        print("N M K")
        print(N, M, K)
        for i in range(N):
            for j in range(M):
                for k in range(K):
                    sliding_window = padded_image[i: i + kernel_size, j: j + kernel_size, k]
                    res[i, j, k] = np.mean(sliding_window)
    return res



"""
matrix = np.array([
    np.array([1,1,1,1,1,1,1]),
    np.array([1,1,1,1,1,1,1]),
    np.array([1,1,1,1,1,1,1])
])

res = pad_image(matrix, 2)
print("res")
print(res)
"""

MATRIX = np.full((2, 3, 7), 5)
res = blur_image(MATRIX, 5)