-
Notifications
You must be signed in to change notification settings - Fork 46
/
Test.py
209 lines (175 loc) · 7.7 KB
/
Test.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
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
#!/usr/bin/env python3
"""
Implementation of RAISR in Python by Jalali-Lab
[RAISR](http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=7744595) (Rapid and Accurate Image Super
Resolution) is an image processing algorithm published by Google Research in 2016. With sufficient
trainingData, consisting of low and high resolution image pairs, RAISR algorithm tries to learn
a set of filters which can be applied to an input image that is not in the training set,
to produce a higher resolution version of it. The source code released here is the Jalali-Lab
implementation of the RAISR algorithm in Python 3.x. The implementation presented here achieved
performance results that are comparable to that presented in Google's research paper
(with ± 0.1 dB in PSNR).
Just-in-time (JIT) compilation employing JIT numba is used to speed up the Python code. A very
parallelized Python code employing multi-processing capabilities is used to speed up the testing
process. The code has been tested on GNU/Linux and Mac OS X 10.13.2 platforms.
Author: Sifeng He, Jalali-Lab, Department of Electrical and Computer Engineering, UCLA
Copyright
---------
RAISR is developed in Jalali Lab at University of California, Los Angeles (UCLA).
More information about the technique can be found in our group website: http://www.photonics.ucla.edu
"""
import numpy as np
import numba as nb
import os
import cv2
import warnings
import pickle
import time
from math import floor
from skimage.measure import compare_psnr
from scipy.misc import imresize
from Functions import *
import matplotlib.image as mpimg
from multiprocessing import Pool
warnings.filterwarnings('ignore')
testPath = 'testData'
R = 2 # Upscaling factor=2 R = [ 2 3 4 ]
patchSize = 11 # Pacth Size=11
gradientSize = 9
Qangle = 24 # Quantization factor of angle =24
Qstrength = 3 # Quantization factor of strength =3
Qcoherence = 3 # Quantization factor of coherence =3
with open("Filters/filter"+str(R), "rb") as fp:
h = pickle.load(fp)
with open("Filters/Qfactor_str"+str(R), "rb") as sp:
stre = pickle.load(sp)
with open("Filters/Qfactor_coh"+str(R), "rb") as cp:
cohe = pickle.load(cp)
filelist = make_dataset(testPath)
wGaussian = Gaussian2d([patchSize, patchSize], 2)
wGaussian = wGaussian/wGaussian.max()
wGaussian = np.diag(wGaussian.ravel())
imagecount = 1
patchMargin = floor(patchSize/2)
psnrRAISR = []
psnrBicubic = []
psnrBlending = []
def TestProcess(i):
if (i < iteration - 1):
offset_i = offset[i * batch:i * batch + batch]
offset2_i = offset2[i * batch:i * batch + batch]
grid = np.tile(gridon[..., None], [1, 1, batch]) + np.tile(offset_i, [patchSize, patchSize, 1])
else:
offset_i = offset[i * batch:im.size]
offset2_i = offset2[i * batch:im.size]
grid = np.tile(gridon[..., None], [1, 1, im.size - (iteration - 1) * batch]) + np.tile(offset_i,[patchSize, patchSize,1])
f = im_LR.ravel()[grid]
gx = im_GX.ravel()[grid]
gy = im_GY.ravel()[grid]
gx = gx.reshape((1, patchSize * patchSize, gx.shape[2]))
gy = gy.reshape((1, patchSize * patchSize, gy.shape[2]))
G = np.vstack((gx, gy))
g1 = np.transpose(G, (2, 0, 1))
g2 = np.transpose(G, (2, 1, 0))
x = Gaussian_Mul(g1, g2, wGaussian)
w, v = np.linalg.eig(x)
idx = (-w).argsort()
w = w[np.arange(np.shape(w)[0])[:, np.newaxis], idx]
v = v[np.arange(np.shape(v)[0])[:, np.newaxis, np.newaxis], np.arange(np.shape(v)[1])[np.newaxis, :, np.newaxis], idx[:,np.newaxis,:]]
thelta = np.arctan(v[:, 1, 0] / v[:, 0, 0])
thelta[thelta < 0] = thelta[thelta < 0] + pi
thelta = np.floor(thelta / (pi / Qangle))
thelta[thelta > Qangle - 1] = Qangle - 1
thelta[thelta < 0] = 0
lamda = w[:, 0]
u = (np.sqrt(w[:, 0]) - np.sqrt(w[:, 1])) / (np.sqrt(w[:, 0]) + np.sqrt(w[:, 1]) + 0.00000000000000001)
lamda = np.searchsorted(stre, lamda)
u = np.searchsorted(cohe, u)
j = thelta * Qstrength * Qcoherence + lamda * Qcoherence + u
j = j.astype('int')
offset2_i = np.unravel_index(offset2_i, (H, W))
t = ((offset2_i[0] - patchMargin) % R) * R + ((offset2_i[1] - patchMargin) % R)
filtertj = h[t, j]
filtertj = filtertj[:, :, np.newaxis]
patch = f.reshape((1, patchSize * patchSize, gx.shape[2]))
patch = np.transpose(patch, (2, 0, 1))
result = np.matmul(patch, filtertj)
return result
print('Begin to process images:')
for image in filelist:
print('\r', end='')
print('' * 60, end='')
print('\r Processing ' + str(imagecount) + '/' + str(len(filelist)) + ' image (' + image + ')')
im_uint8 = cv2.imread(image)
im_mp = mpimg.imread(image)
if len(im_mp.shape) == 2:
im_uint8 = im_uint8[:,:,0]
im_uint8 = modcrop(im_uint8, R)
if len(im_uint8.shape) > 2:
im_ycbcr = BGR2YCbCr(im_uint8)
im = im_ycbcr[:, :, 0]
else:
im = im_uint8
im_double = im2double(im)
H, W = im.shape
region = (slice(patchMargin, H - patchMargin), slice(patchMargin, W - patchMargin))
start = time.time()
imL = imresize(im_double, 1 / R, interp='bicubic', mode='F')
im_bicubic = imresize(imL, (H, W), interp='bicubic', mode='F')
im_bicubic = im_bicubic.astype('float64')
im_bicubic = np.clip(im_bicubic, 0, 1)
im_LR = np.zeros((H+patchSize-1,W+patchSize-1))
im_LR[(patchMargin):(H+patchMargin),(patchMargin):(W+patchMargin)] = im_bicubic
im_result = np.zeros((H, W))
im_GX, im_GY = np.gradient(im_LR)
index = np.array(range(im_LR.size)).reshape(im_LR.shape)
offset = np.array(index[0:H, 0:W].ravel())
offset2 = np.array(range(im.size))
gridon = index[0:patchSize, 0:patchSize]
batch = 2000
iteration = ceil(im.size / batch + 0.000000000000001)
im_result = np.array([])
i = range(iteration)
p = Pool()
im_in = p.map(TestProcess, i)
for i in range(iteration):
im_result = np.append(im_result, im_in[i])
im_result = im_result.reshape(H, W)
im_result = np.clip(im_result, 0, 1)
end = time.time()
print(end - start)
im_blending = Blending2(im_bicubic, im_result)
# im_blending = Backprojection(imL, im_blending, 50) #Optional: Backprojection, which can slightly improve PSNR, especilly for large upscaling factor.
im_blending = np.clip(im_blending, 0, 1)
if len(im_uint8.shape) > 2:
result_ycbcr = np.zeros((H, W, 3))
result_ycbcr[:, :, 0] = im_blending * 255
result_ycbcr[:, :, 1] = Prepare(im_ycbcr[:, :, 1], patchSize, R) * 255
result_ycbcr[:, :, 2] = Prepare(im_ycbcr[:, :, 2], patchSize, R) * 255
result_ycbcr = result_ycbcr[region].astype('uint8')
result_RAISR = YCbCr2BGR(result_ycbcr)
else:
result_RAISR = im_result[region] * 255
result_RAISR = result_RAISR.astype('uint8')
im_result = im_result*255
im_result = np.rint(im_result).astype('uint8')
im_bicubic = im_bicubic*255
im_bicubic = np.rint(im_bicubic).astype('uint8')
im_blending = im_blending * 255
im_blending = np.rint(im_blending).astype('uint8')
PSNR_bicubic = compare_psnr(im[region], im_bicubic[region])
PSNR_result = compare_psnr(im[region], im_result[region])
PSNR_blending = compare_psnr(im[region], im_blending[region])
PSNR_blending = max(PSNR_result, PSNR_blending)
createFolder('./results/')
cv2.imwrite('results/' + os.path.splitext(os.path.basename(image))[0] + '_result.bmp', result_RAISR)
psnrRAISR.append(PSNR_result)
psnrBicubic.append(PSNR_bicubic)
psnrBlending.append(PSNR_blending)
imagecount += 1
RAISR_psnrmean = np.array(psnrBlending).mean()
Bicubic_psnrmean = np.array(psnrBicubic).mean()
print('\r', end='')
print('' * 60, end='')
print('\r RAISR PSNR of '+ testPath +' is ' + str(RAISR_psnrmean))
print('\r Bicubic PSNR of '+ testPath +' is ' + str(Bicubic_psnrmean))