-
Notifications
You must be signed in to change notification settings - Fork 2
/
GPStatistics.py
154 lines (112 loc) · 7.71 KB
/
GPStatistics.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
from scipy import math
from cv2.ximgproc import guidedFilter
from skimage.restoration import denoise_tv_chambolle
from operator import mul, sub, add
import numpy as np
from skimage.io import imread, imsave
def protectedSqrt(root):
return math.sqrt(abs(root))
def protectedDiv(left, right):
if (right != 0):
return left / right
else:
return 1
def Max2(a,b):
return max(a,b)
def Min2(a,b):
return min(a,b)
def AbsSum(a,b):
return abs(a+b)
def AbsSub(a,b):
return abs(a-b)
def pow2(a):
return pow(a,2)
def protectedLog(a):
if(a <= 0):
return 1
else:
return math.log(a)
def protectedExp(a):
if(a >= 255):
a = 1
return math.exp(a)
else:
a = a/255.0
return math.exp(a)
def myif(a,b,c,d):
if(a>=b):
return c
else:
return d
def estimateAirlight(degraded_I, winSize):
Nr,Nc,Np = degraded_I.shape
I1 = np.concatenate((np.flipud(np.fliplr(degraded_I)), np.flipud(degraded_I), np.flipud(np.fliplr(degraded_I))), axis=1)
I2 = np.concatenate((np.fliplr(degraded_I), degraded_I, np.fliplr(degraded_I)), axis=1)
I3 = np.concatenate((np.flipud(np.fliplr(degraded_I)), np.flipud(degraded_I), np.flipud(np.fliplr(degraded_I))), axis=1)
padded_I = np.concatenate( (I1,I2,I3) ,axis=0)
padded_I = np.double( padded_I[ int(Nr-((winSize-1)/2.0)):int(2.0*Nr+((winSize-1)/2.0)), int(Nc-((winSize-1)/2.0)):int(2.0*Nc+((winSize-1)/2.0)),: ] )
estimate_A = np.zeros([Nr,Nc])
f_mv = np.zeros([winSize,winSize])
for k in range(Nr):
for l in range(Nc):
f_mv = ( padded_I[ k:winSize+k, l:winSize+l, : ] )
f_max = f_mv.max()
f_min = f_mv.min()
u = (f_min + f_max) / 2.0
v = f_max - f_min
estimate_A[k,l] = u / (1.0 + v)
x0,y0 = np.where( estimate_A == estimate_A.max())
A = np.zeros(3)
A[0] = degraded_I[x0[0], y0[0],0]
A[1] = degraded_I[x0[0], y0[0],1]
A[2] = degraded_I[x0[0], y0[0],2]
estimated_A = 0.333*A[0] + 0.333*A[1] + 0.333*A[2]
print('Estimate Airlight: Done!')
return estimated_A
###############################################################################
def dehazeScene(degraded_I, A_est, winSize):
Nr,Nc,Np = degraded_I.shape
I1 = np.concatenate((np.flipud(np.fliplr(degraded_I)), np.flipud(degraded_I), np.flipud(np.fliplr(degraded_I))), axis=1)
I2 = np.concatenate((np.fliplr(degraded_I), degraded_I, np.fliplr(degraded_I)), axis=1)
I3 = np.concatenate((np.flipud(np.fliplr(degraded_I)), np.flipud(degraded_I), np.flipud(np.fliplr(degraded_I))), axis=1)
padded_I = np.concatenate( (I1,I2,I3) ,axis=0 )
padded_I = np.double( padded_I[ int(Nr-((winSize-1)/2.0)):int(2.0*Nr+((winSize-1)/2.0)), int(Nc-((winSize-1)/2.0)):int(2.0*Nc+((winSize-1)/2.0)),: ] )
estimated_T = np.zeros([Nr,Nc])
refined_T = np.zeros([Nr,Nc])
dehazed_I = np.zeros([Nr,Nc,Np])
for k in range(Nr):
for l in range(Nc):
f_v = padded_I[ k:winSize+k, l:winSize+l, : ]
MAX = np.max(f_v)
MIN = np.min(f_v)
RANGE = (MAX - MIN)
MRANGE = (MAX + MIN)/2.0
MEAN = np.mean(f_v)
VAR = np.var(f_v)
ALPHA = RANGE/(A_est)
BETA = 1 - ALPHA
#Estimator GP-SD
# estimated_T[k,l] = protectedExp(add(sub(myif(VAR, AbsSub(RANGE, myif(protectedDiv(protectedExp(add(VAR, 229.5)), protectedDiv(mul(VAR, RANGE), mul(MEAN, MAX))), myif(add(abs(RANGE), protectedSqrt(MEAN)), AbsSum(add(RANGE, sub(MAX, 229.5)), protectedLog(MEAN)), mul(mul(mul(MAX, MIN), MRANGE), Min2(MIN, MRANGE)), add(Min2(ALPHA, VAR), protectedSqrt(VAR))), sub(sub(VAR, abs(ALPHA)), RANGE), add(add(add(AbsSum(MAX, ALPHA), protectedLog(myif(MAX, MAX, sub(MAX, myif(sub(RANGE, RANGE), Max2(BETA, MIN), mul(MIN, MRANGE), Min2(VAR, MRANGE))), add(229.5, abs(AbsSub(mul(MAX, ALPHA), add(MAX, ALPHA))))))), BETA), ALPHA))), MRANGE, myif(sub(RANGE, MAX), Max2(BETA, MIN), mul(MIN, MRANGE), Min2(VAR, MRANGE))), MRANGE), add(sub(RANGE, MAX), protectedLog(myif(protectedDiv(AbsSub(sub(RANGE, MEAN), abs(MIN)), add(sub(RANGE, MAX), protectedLog(myif(protectedDiv(229.5, sub(sub(protectedExp(229.5), Min2(MRANGE, MAX)), Min2(MIN, MRANGE))), myif(protectedDiv(229.5, MAX), protectedLog(MAX), MRANGE, add(MAX, protectedLog(RANGE))), protectedLog(pow2(MAX)), MAX)))), myif(protectedDiv(229.5, 229.5), protectedLog(MAX), sub(sub(VAR, abs(pow2(229.5))), RANGE), add(add(add(sub(RANGE, Max2(MRANGE, MEAN)), protectedLog(myif(MAX, MAX, sub(MAX, sub(RANGE, MAX)), add(229.5, VAR)))), sub(myif(VAR, AbsSub(AbsSub(sub(229.5, MEAN), abs(MIN)), 229.5), MRANGE, myif(sub(RANGE, MAX), Max2(protectedLog(Min2(protectedExp(MRANGE), Min2(VAR, MAX))), MIN), mul(MIN, MRANGE), 229.5)), AbsSub(sub(RANGE, MEAN), abs(MIN)))), ALPHA)), protectedLog(protectedDiv(Max2(MRANGE, RANGE), sub(MRANGE, protectedDiv(AbsSub(MAX, 229.5), pow2(Max2(RANGE, MIN)))))), myif(add(abs(RANGE), add(RANGE, MRANGE)), AbsSum(add(RANGE, ALPHA), protectedLog(MEAN)), mul(mul(MIN, MRANGE), Min2(MIN, MRANGE)), add(Min2(ALPHA, VAR), protectedSqrt(VAR))))))))
#Estimator GP-SN
estimated_T[k,l] = protectedDiv(protectedExp(myif(229.5, MIN, sub(Min2(MRANGE, VAR), AbsSum(sub(Min2(BETA, add(add(protectedDiv(ALPHA, MRANGE), sub(MIN, MIN)), sub(add(protectedDiv(RANGE, RANGE), AbsSub(RANGE, protectedDiv(pow2(RANGE), protectedExp(AbsSum(Min2(Min2(MEAN, VAR), mul(VAR, VAR)), Min2(pow2(229.5), Min2(229.5, ALPHA))))))), MIN))), MIN), protectedSqrt(sub(Min2(Min2(pow2(229.5), Min2(229.5, ALPHA)), add(add(sub(add(protectedDiv(MEAN, VAR), MEAN), MIN), AbsSub(RANGE, ALPHA)), sub(add(protectedDiv(MIN, Min2(MEAN, VAR)), AbsSub(RANGE, protectedDiv(pow2(RANGE), protectedExp(mul(myif(229.5, 229.5, RANGE, 229.5), protectedDiv(ALPHA, ALPHA)))))), VAR))), MIN)))), add(add(protectedDiv(ALPHA, sub(Min2(Min2(pow2(229.5), Min2(229.5, ALPHA)), add(add(sub(add(protectedDiv(MEAN, VAR), MEAN), MIN), AbsSub(RANGE, ALPHA)), sub(add(protectedDiv(MIN, VAR), AbsSub(RANGE, protectedDiv(pow2(RANGE), protectedExp(mul(myif(229.5, 229.5, RANGE, 229.5), protectedDiv(ALPHA, ALPHA)))))), add(protectedDiv(ALPHA, MRANGE), sub(MIN, MIN))))), MIN)), sub(MIN, MIN)), sub(add(protectedDiv(MIN, ALPHA), AbsSub(RANGE, protectedDiv(pow2(RANGE), protectedExp(MEAN)))), MIN)))), protectedExp(MEAN))
print('Estimate Transmission: Done!')
estimated_T[estimated_T > 1] = 1.0
estimated_T[estimated_T < 0.1] = 0.1
estimated_T = guidedFilter(np.uint8(degraded_I*255), np.uint8(estimated_T*255), 100, 0.5)/255.0
refined_T = denoise_tv_chambolle(estimated_T, weight=0.1, multichannel=False)
refined_T[refined_T > 1] = 1
refined_T[refined_T < 0.1] = 0.1
dehazed_I[:,:,0] = ( (degraded_I[:,:,0] - A_est) / refined_T ) + A_est
dehazed_I[:,:,1] = ( (degraded_I[:,:,1] - A_est) / refined_T ) + A_est
dehazed_I[:,:,2] = ( (degraded_I[:,:,2] - A_est) / refined_T ) + A_est
dehazed_I[dehazed_I > 255.0] = 255.0
dehazed_I[dehazed_I < 0.0] = 0.0
print('Estimate Dehazed Image: Done!')
return dehazed_I
if __name__ == "__main__":
imgName = 'img7'
degradedScene = np.float64( imread('Images/%s.png' % (imgName) ) ) # misc.imread('Images/%s.png' % (imgName)) )
degradedScene = degradedScene[:,:,0:3]
estimatedA = estimateAirlight(degradedScene, 19)
dehazed_Img = dehazeScene(degradedScene, estimatedA, 5)
imsave('Results/%s_outputSN.png' % (imgName), np.uint8(dehazed_Img))