本文用 Python 实现 PS 滤镜中的旋转模糊，具体的算法原理和效果可以参考之前的博客：

http://blog.csdn.net/matrix_space/article/details/39211667

from skimage import img_as_float

import matplotlib.pyplot as plt

from skimage import io

import numpy as np

import numpy.matlib

file_name='D:/Visual Effects/PS Algorithm/4.jpg'

img=io.imread(file_name)

img = img_as_float(img)

img_out = img.copy()

row, col, channel = img.shape

xx = np.arange (col)

yy = np.arange (row)

x_mask = numpy.matlib.repmat (xx, row, 1)

y_mask = numpy.matlib.repmat (yy, col, 1)

y_mask = np.transpose(y_mask)

center_y = (row -1) / 2.0

center_x = (col -1) / 2.0

R = np.sqrt((x_mask - center_x) **2 + (y_mask - center_y) ** 2)

angle = np.arctan2(y_mask - center_y , x_mask - center_x)

Num = 20

arr = ( np.arange(Num) + 1 ) / 100.0

for i in range (row):

    for j in range (col):

        T_angle = angle[i, j] + arr

        new_x = R[i, j] * np.cos(T_angle) + center_x

        new_y = R[i, j] * np.sin(T_angle) + center_y

        int_x = new_x.astype(int)

        int_y = new_y.astype(int)

        int_x[int_x > col-1] = col - 1

        int_x[int_x < 0] = 0

        int_y[int_y < 0] = 0

        int_y[int_y > row -1] = row -1

        img_out[i,j,0] = img[int_y, int_x, 0].sum()/Num

        img_out[i,j,1] = img[int_y, int_x, 1].sum()/Num

        img_out[i,j,2] = img[int_y, int_x, 2].sum()/Num

plt.figure(1)

plt.imshow(img)

plt.axis('off')

plt.figure(2)

plt.imshow(img_out)

plt.axis('off')

plt.show()