-
Notifications
You must be signed in to change notification settings - Fork 10
/
compute_psnr.py
136 lines (120 loc) · 4.28 KB
/
compute_psnr.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
import argparse
import cv2
import os
import numpy as np
from skimage.metrics import mean_squared_error
from skimage.measure import compare_ssim
from skimage.metrics import structural_similarity
from skimage.metrics import peak_signal_noise_ratio
#import lpips
import torch
from tqdm import tqdm
#from niqe.niqe import compute_niqe
#criterion = lpips.LPIPS(net='vgg', lpips=True, pnet_rand=False, pretrained=True).cuda()
def rgb2ycbcr(im, only_y=True):
'''
same as matlab rgb2ycbcr
:parame img: uint8 or float ndarray
'''
in_im_type = im.dtype
im = im.astype(np.float64)
if in_im_type != np.uint8:
im *= 255.
# convert
if only_y:
rlt = np.dot(im, np.array([65.481, 128.553, 24.966])/ 255.0) + 16.0
else:
rlt = np.matmul(im, np.array([[65.481, -37.797, 112.0 ],
[128.553, -74.203, -93.786],
[24.966, 112.0, -18.214]])/255.0) + [16, 128, 128]
if in_im_type == np.uint8:
rlt = rlt.round()
else:
rlt /= 255.
return rlt.astype(in_im_type)
def rgb2ycbcrTorch(im, only_y=True):
'''
same as matlab rgb2ycbcr
Input:
im: float [0,1], N x 3 x H x W
only_y: only return Y channel
'''
im_temp = im.permute([0,2,3,1]) * 255.0 # N x H x W x C --> N x H x W x C, [0,255]
# convert
if only_y:
rlt = torch.matmul(im_temp, torch.tensor([65.481, 128.553, 24.966],
device=im.device, dtype=im.dtype).view([3,1])/ 255.0) + 16.0
else:
rlt = torch.matmul(im_temp, torch.tensor([[65.481, -37.797, 112.0 ],
[128.553, -74.203, -93.786],
[24.966, 112.0, -18.214]],
device=im.device, dtype=im.dtype)/255.0) + \
torch.tensor([16, 128, 128]).view([-1, 1, 1, 3])
rlt /= 255.0
rlt.clamp_(0.0, 1.0)
return rlt.permute([0, 3, 1, 2])
def readim(file):
# print(file)
img = cv2.imread(file)
img = img.astype(np.float32)
return img / 255.
def loadfiles(folder):
files = os.listdir(folder)
return natsorted(files)
def resize(im, size, crop=True):
if crop:
return im[:size[0], :size[1]]
else:
return cv2.resize(im, size)
from natsort import natsorted
def np2torch(img):
im = img.astype(np.float32) / 255
im = torch.tensor(im).permute((2,0,1)).unsqueeze(0)
return im.cuda()
def compute_metrics(path1, path2, ycbcr=True):
print(path1)
files1 = loadfiles(path1)
files2 = loadfiles(path2)
print(len(files1), len(files2))
psnr = []
ssim = []
mse = []
lpips = []
niqe = []
crop = False
for file1, file2 in tqdm(zip(files1, files2)):
img1 = readim(os.path.join(path1, file1))
img2 = readim(os.path.join(path2, file2))
if img1.shape != img2.shape:
if not crop:
img1 = resize(img1, img2.shape[:2][::-1], False)
else:
img1 = resize(img1, img2.shape, True)
# print(img1.shape, img2.shape, img1.max())
MSE = mean_squared_error(img1, img2)
if ycbcr:
img1 = rgb2ycbcr(img1, True)
img2 = rgb2ycbcr(img2, True)
diff = (img2 - img1)
# print(diff.mean(), diff.max(), diff.min(), diff.shape)
PSNR = peak_signal_noise_ratio(img1, img2, data_range=1)
SSIM = structural_similarity(img1, img2, win_size=11, multichannel=False if ycbcr else True, data_range=1)
mse.append(MSE)
psnr.append(PSNR)
ssim.append(SSIM)
mean_mse, mean_psnr, mean_ssim = np.mean(mse), np.mean(psnr), np.mean(ssim)
print(mean_mse, mean_psnr, mean_ssim)
return mean_mse, mean_psnr, mean_ssim
if __name__ == "__main__":
parser = argparse.ArgumentParser()
# path setting
parser.add_argument('--path1', type=str,default= "") # modify the experiments name-->modify all save path
parser.add_argument('--path2', type=str,default= "")
args = parser.parse_args()
path1 = ''
path2 = ''
if len(args.path1) != 0:
path1 = args.path1
if len(args.path2) != 0:
path2 = args.path2
compute_metrics(path1, path2, True)