/
weiner.py
executable file
·159 lines (119 loc) · 4.35 KB
/
weiner.py
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
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
"""
=====================
Image Deconvolution
=====================
In this example, we deconvolve a noisy version of an image using Wiener
and unsupervised Wiener algorithms. This algorithms are based on
linear models that can't restore sharp edge as much as non-linear
methods (like TV restoration) but are much faster.
"""
import numpy as np
import matplotlib.pyplot as plt
from skimage import color, data, restoration
import cv2
from scipy.signal import convolve2d as conv2
import numpy as np
from numpy.fft import fft, ifft, ifftshift
"""
implementination of Wiener deconvolution including the signal to noise ratio
numpy implementation does not have snr parameter.
"""
def wiener_deconv(signal, kernel, snr):
kernel = np.hstack((kernel, np.zeros(len(signal) - len(kernel))))
H = fft(kernel)
deconv = np.real(ifft (fft (signal)* np.conj(H) / (H * np.conj(H) + snr**2)))
return deconv
def weiner_filter(img,psf):
astro = color.rgb2gray(data.astronaut())
print "running Weiner for a channel"
out = np.ndarray(shape=img.shape, dtype=np.float64) # redifine to correct type
out = np.copy(img)
min_intensity = np.amin(img)
out = out - min_intensity
max_intensity = np.amax(out)
out = out * (1.0 / max_intensity)
img = out
#psf = np.ones((5, 5)) / 25.0
astro = conv2(astro, psf, 'same')
astro += 0.22 * astro.std() * np.random.standard_normal(astro.shape)
#deconvolved, _ = restoration.unsupervised_wiener(img, psf)
#deconvolved = restoration.wiener(img, psf, 500)
deconvolved = restoration.richardson_lucy(img, psf, 200)
return deconvolved
def generate_PSF(kernlen=21, nsig=3):
import numpy as np
import scipy.ndimage.filters as fi
# create nxn zeros
inp = np.zeros((kernlen, kernlen))
# set element at the middle to one, a dirac delta
inp[kernlen//2, kernlen//2] = 1
# gaussian-smooth the dirac, resulting in a gaussian filter mask
value = fi.gaussian_filter(inp, nsig)
#print value
return value
def run_weiner(outimg,psf):
image = '/home/dpant/CP/FinalProject/images_self/ab.png'
image = "/home/dpant/CP/FinalProject/text.png"
img_org = cv2.imread(image)
img = img_org
num_channels = img.shape[2]
#img = cv2.cvtColor(img_org, cv2.COLOR_BGR2GRAY)
#hdr_image = np.zeros(images[0].shape, dtype=np.float64)
final_image = np.zeros(img.shape, dtype=np.float64)
#img = restoration.denoise_bilateral(img)
#(img, sigma_color=0.05, sigma_spatial=15)
xyz = psf * 255.0
cv2.imwrite("denoise.png",xyz)
#psf = fspecial(8,img.shape[0])
#print np.sum(psf)
#import sys
#sys.exit()
for channel in range(num_channels):
# collect the current layer of each input image
layer_stack = img[:, :, channel]
#layer_stack = [img[:, :, channel] for img in images]
channel_res = weiner_filter(layer_stack,psf)
min = np.min(channel_res)
channel_res -= min
max = np.max(channel_res)
channel_res = channel_res/max
channel_res *= 255.0
# Sample image intensities
#intensity_samples = sampleIntensities(layer_stack, num_points)
final_image[..., channel] = channel_res
blur_out = cv2.GaussianBlur(final_image, (5,5), 2)
final_image = cv2.addWeighted(final_image, 1.5, blur_out, -0.5, 0);
cv2.imwrite(outimg,final_image)
np.set_printoptions(precision=3,threshold=1000,linewidth=1000)
#threshold=None, edgeitems=None,
def fspecial(rad,mat_size=0):
if not mat_size:
mat_size = (int) (2* rad + 1)
def getnc(x):
return x - mat_size/2
mat = np.zeros( (mat_size,mat_size) )
for i in range(mat_size):
for j in range(mat_size):
x = getnc(i)
y = getnc(j)
r = np.sqrt(x*x + y*y)
#print r , rad
#if(r<=rad):
# print x, y
mat[i,j] = (r <= rad)
print mat
mat = mat/np.sum(mat)
print mat
print mat.shape
return mat
if __name__ == '__main__':
psf = np.loadtxt('psf8.txt', dtype=np.float, comments='#', delimiter=',')
#psf = np.ones((17, 17)) / (17.0 * 17.0)
v = generate_PSF(5,0.5)
#print np.sum(v)
for cnt in np.arange(1,19,1.0):
#psf = generate_PSF(cnt,0.5)
#psf = np.ones((cnt, cnt)) / (cnt * cnt * 1.0)
outimg = "final_" + str(cnt) + ".png"
psf = fspecial(cnt)
run_weiner(outimg,psf)