forked from AdamKaniasty/SVD-Image-Compression
-
Notifications
You must be signed in to change notification settings - Fork 0
/
app.py
120 lines (101 loc) · 3.68 KB
/
app.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
import os
from PIL import Image
import matplotlib.pyplot as plt
import numpy as np
from tqdm import tqdm
path_grey = './images/gray'
path_color = './images/color'
color_images = []
gray_images = []
color_images_names = []
gray_images_names = []
for filename in os.listdir(path_color):
if filename.endswith('.jpg'):
img = Image.open(os.path.join(path_color, filename))
img_matrix = np.array(img)
color_images.append(img_matrix)
color_images_names.append(filename[:-4])
for filename in os.listdir(path_grey):
if filename.endswith('.jpg'):
img = Image.open(os.path.join(path_grey, filename)).convert('L')
img_matrix = np.array(img)
gray_images.append(img_matrix)
gray_images_names.append(filename[:-4])
def performSVD(img, gray=True, k=0.05):
num = None
if gray:
U, S, V = np.linalg.svd(img, full_matrices=False)
num = round(S.shape[0] * k)
return [U[:, :num], S[:num], V[:num, :], num]
else:
U = []
S = []
V = []
for i in range(3):
U_temp, S_temp, V_temp = np.linalg.svd(img[:, :, i])
num = round(S_temp.shape[0] * k)
U.append(U_temp[:, :num])
S.append(S_temp[:num])
V.append(V_temp[:num, :])
return [U, S, V, num]
def reconstructSVD(U, S, V, gray=True):
if gray:
return np.matrix(U) * np.diag(S) * np.matrix(V)
else:
print(U[0].shape, S[0].shape, V[0].shape)
svd_r = reconstructSVD(U[0], S[0], V[0], True)
svd_g = reconstructSVD(U[1], S[1], V[1], True)
svd_b = reconstructSVD(U[2], S[2], V[2], True)
img = np.zeros((U[0].shape[0], V[0].shape[1], 3))
img[:, :, 1] = svd_g
img[:, :, 0] = svd_r
img[:, :, 2] = svd_b
return img
def spaceSavings(imgarr, k, gray=True):
rows, cols = imgarr.shape[:2]
bpp = imgarr.itemsize * 8
bps = np.dtype(np.float32).itemsize * 8
if gray:
orig_size = rows * cols * bpp
rankk_size = (rows * k + k + k * cols) * bps
saved = (orig_size - rankk_size) / orig_size * 100
else:
orig_size = rows * cols * 3 * bpp
rankk_size = (rows * k + k + k * cols) * 3 * bps
saved = (orig_size - rankk_size) / orig_size * 100
return saved
def processIMG(img, name, k, gray=True):
svd = performSVD(img, gray, k)
num = svd[3]
reconstruction = reconstructSVD(svd[0], svd[1], svd[2], gray)
plt.figure(figsize=(15, 6))
saved = spaceSavings(img, num, gray)
plt.subplot(121)
if gray:
plt.imshow(img, cmap='gray')
plt.subplot(122)
plt.imshow(reconstruction, cmap='gray')
plt.title(f'{name}, k = {k}, saved={saved:.2f}%')
plt.savefig(f'./images/gray/figs/{name}_{k}.jpg')
plt.close()
plt.imsave(f'./images/gray/reconstructed/{name}_{k}.jpg', reconstruction, cmap='gray')
else:
plt.imshow(img)
plt.subplot(122)
plt.imshow(reconstruction.astype(np.uint8))
plt.title(f'{name}, k = {k}, saved={saved:.2f}%')
plt.savefig(f'./images/color/figs/{name}_{k}.jpg')
plt.close()
plt.imsave(f'./images/color/reconstructed/{name}_{k}.jpg', reconstruction.astype(np.uint8))
def testPerformanceGray():
k_params = [0.01, 0.05, 0.1, 0.2, 0.5]
for (index, img) in tqdm(enumerate(gray_images)):
for k in tqdm(k_params):
processIMG(img, gray_images_names[index], k)
def testPerformanceColor():
k_params = [0.01, 0.05, 0.1, 0.2, 0.5]
for (index, img) in tqdm(enumerate(color_images)):
for k in tqdm(k_params):
processIMG(img, color_images_names[index], k, False)
testPerformanceGray()
# testPerformanceColor()