图像/高斯滤波器阵列上的核矩阵3x3

Question

我正在尝试实现一个高斯滤波器。为此，我使用一个内核3x3和一个image.The问题的数组，为数组的每个i，j元素定义一个子矩阵3x3。我在代码中写了详细信息。

import numpy as np
import scipy
from scipy import misc
import matplotlib.pyplot as plt


imagen = scipy.misc.imread("C:\\Users\\Reymi\\Downloads\\imagen.png") #importing image of original size (1929, 1280)

imagen_real = scipy.pad(array=imagen, pad_width=[1, 1], mode='constant', constant_values=0) #add 1 column and 1 row of zeros to avoid the kernel of going outside the array. size is (1931, 1282)

imagen_nueva = np.empty((1931, 1282)) #the new image. the same size as the image I will filter

(dim_x,dim_y)=np.shape(imagen_real)


ker1 = np.array([[1/16, 1/8, 1/16],   #3x3 kernel
                [1/8, 1/4, 1/8],
                [1/16, 1/8, 1/16]])

def multiplicar_entero():

    global imagen_nueva
    for i in range(1,dim_x): #the range starts from 1 to avoid the column and row of zeros, and ends before the last col and row of zeros
        for j in range(1,dim_y):
            imagen_entry = np.empty((3, 3))      #Main problem here: how do I define a 3x3 matrix for each entry?
            valor = np.sum(imagen_entry*ker1)    #Matrix 3x3 is filled with the elements around each [i, j] entry of the array
            imagen_real[i, j] = valor
            imagen_nueva = np.append(imagen_real[i, j], (1931, 1282)) #This is supposed to each new [i, j] entry to the new image

print("La imagen con el blur effect es la siguiente:\n")

multiplicar_entero()   #calls function


plt.imshow(imagen_nueva)  #Show new image
plt.gray()
plt.show()

抱歉密码太长了。谢谢你的帮助。

Answer 1

您的代码有几个问题如下：

• 你使用的是一个我们无法访问的图像。在示例中始终使用免费可用的图像，这样我们就可以运行代码。我这里用的是scipy.misc.ascent。
• 除非必须使用全局变量，否则不要使用全局变量。
• 始终用英文编码，并带有英文变量名和英文注释。这样的评论就容易多了。

我在下面对您的代码做了一些修正，我想您可以看到我是如何解决问题的。具体来说，您希望使用:索引，它允许提取数组的子集并使用array[i, j]分配结果：

import numpy as np
import scipy
from scipy import misc
import matplotlib.pyplot as plt

imagen = scipy.misc.ascent()  # Freely available image

(dim_x, dim_y) = np.shape(imagen)

ker1 = np.array([[1/16, 1/8, 1/16],   #3x3 kernel
                [1/8, 1/4, 1/8],
                [1/16, 1/8, 1/16]])

def multiplicar_entero(imagen):
    imagen_nueva = np.zeros(imagen.shape) #the new image. the same size as the image I will filter
    for i in range(1,dim_x-1): #the range starts from 1 to avoid the column and row of zeros, and ends before the last col and row of zeros
        for j in range(1,dim_y-1):
            imagen_entry = imagen[i-1:i+2, j-1:j+2]
            valor = np.sum(imagen_entry*ker1)    #Matrix 3x3 is filled with the elements around each [i, j] entry of the array
            imagen_nueva[i, j] = valor
    return imagen_nueva

imagen_nueva = multiplicar_entero(imagen)

plt.imshow(imagen_nueva)  #Show new image
plt.gray()
plt.show()