使用numpy/scipy对矩阵堆栈中的每一层进行卷积

import numpy as np
from scipy.signal import convolve

# 定义输入矩阵堆栈
stack = np.array([[[1, 2, 3],
                   [4, 5, 6],
                   [7, 8, 9]],
                  
                  [[10, 11, 12],
                   [13, 14, 15],
                   [16, 17, 18]],
                  
                  [[19, 20, 21],
                   [22, 23, 24],
                   [25, 26, 27]]])

# 定义卷积核
kernel = np.array([[1, 0, -1],
                   [2, 0, -2],
                   [1, 0, -1]])

# 对矩阵堆栈中的每一层进行卷积
result = np.zeros_like(stack)
for i in range(stack.shape[0]):
    result[i] = convolve(stack[i], kernel, mode='same')

print(result)