-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathpreprocess_image_files.py
executable file
·190 lines (155 loc) · 6.5 KB
/
preprocess_image_files.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
#! /usr/bin/env python3
import argparse
import os
import skimage
import numpy as np
import math
import matplotlib
import PIL
from skimage import io, filters, transform, util, draw, exposure
from PIL import Image, ImageDraw, ImageFont
parser = argparse.ArgumentParser("Preprocess image files.")
parser.add_argument("--dir",
required=True,
help="directory of images to process")
args = parser.parse_args()
img_dir = args.dir
print("process images in ", img_dir)
os.chdir(path=img_dir)
orig_dir = "original"
blur_dir = "blurred"
compr_dir = "compressed"
cropped_dir = "cropped"
decimated_dir = "decimated"
upscale_dir = "upscale"
downscale_dir = "downscale"
noise_dir = "noise"
occluded_dir = "occluded"
rotated_dir = "rotated"
horizflip_dir = "hflip"
vertflip_dir = "vflip"
brighten_dir = "brighten"
darken_dir = "darken"
histeq_dir = "histeq"
shear_dir = "shear"
try:
print("Create Directories")
os.mkdir(orig_dir)
os.mkdir(blur_dir)
os.mkdir(compr_dir)
os.mkdir(cropped_dir)
os.mkdir(decimated_dir)
os.mkdir(noise_dir)
os.mkdir(occluded_dir)
os.mkdir(rotated_dir)
os.mkdir(upscale_dir)
os.mkdir(downscale_dir)
os.mkdir(horizflip_dir)
os.mkdir(vertflip_dir)
os.mkdir(brighten_dir)
os.mkdir(darken_dir)
os.mkdir(histeq_dir)
os.mkdir(shear_dir)
except FileExistsError:
print("Some Directories already exist.")
def blur_image_and_save(file, img, sigma):
path = os.path.join(img_dir, blur_dir, file)
blurred_img = filters.gaussian(img, sigma=sigma)
io.imsave(path, blurred_img, plugin="pil", quality=100)
def compress_image_and_save(file, img, quality=50):
path = os.path.join(img_dir, compr_dir, file)
io.imsave(path, img, plugin="pil", quality=quality)
def crop_image_and_save(file, img, pct_margin=0.05):
path = os.path.join(img_dir, cropped_dir, file)
h, w, d = img.shape
w_margin = int(w*pct_margin)
h_margin = int(h*pct_margin)
cropped = util.crop(img, ((h_margin, h_margin), (w_margin, w_margin),(0,0)), copy=True)
io.imsave(path, cropped, plugin="pil", quality=100)
def decimate_image_and_save(file, img, factors):
path = os.path.join(img_dir, decimated_dir, file)
decimated = transform.downscale_local_mean(img, factors)
decimated = exposure.rescale_intensity(decimated, in_range='image', out_range=(0.0,1.0))
decimated = util.img_as_ubyte(decimated)
io.imsave(path, decimated, plugin="pil", quality=100)
def scale_image_and_save(file, img, dir_name, scale_factor):
path = os.path.join(img_dir, dir_name, file)
scaled = transform.rescale(img, scale_factor, anti_aliasing_sigma=1.0)
io.imsave(path, scaled, plugin="pil", quality=100)
def noise_image_and_save(file, img, gaussian_var=0.005):
path = os.path.join(img_dir, noise_dir, file)
noisy = util.random_noise(img, mode='gaussian', mean=0, var=gaussian_var)
io.imsave(path, noisy, plugin="pil", quality=100)
def occlude_image_and_save(file, img, font_size=40):
path = os.path.join(img_dir, occluded_dir, file)
h, w, d = img.shape
y = int(0.75*h)
x = int(0.10*w)
myfont = ImageFont.truetype("/usr/share/fonts/truetype/dejavu/DejaVuSans.ttf", font_size)
imgpil = Image.fromarray(img)
draw_ctx = ImageDraw.Draw(imgpil)
draw_ctx.text((x,y), file, fill=(255, 255, 0), font=myfont)
occluded_img = np.asarray(imgpil)
io.imsave(path, occluded_img, plugin="pil", quality=100)
def rotate_image_and_save(file, img, degrees=4.0):
path = os.path.join(img_dir, rotated_dir, file)
rotd_img = transform.rotate(img, degrees, resize=False)
h, w, d = rotd_img.shape
h_margin = int(0.05 * h)
w_margin = int(0.05 * w)
cropped_img = util.crop(rotd_img, ((h_margin, h_margin), (w_margin, w_margin), (0, 0)), copy=True)
io.imsave(path, cropped_img, plugin="pil", quality=100)
def hflip_image_and_save(file, img):
path = os.path.join(img_dir, horizflip_dir, file)
flipped = np.fliplr(img)
io.imsave(path, flipped, plugin="pil", quality=100)
def vflip_image_and_save(file, img):
path = os.path.join(img_dir, vertflip_dir, file)
flipped = np.flipud(img)
io.imsave(path, flipped, plugin="pil", quality=100)
def gamma_correction_and_save(file, img, dir_name, gamma=1):
path = os.path.join(img_dir, dir_name, file)
imgfl = util.img_as_float(img)
img2 = exposure.adjust_gamma(imgfl, gamma=gamma)
img2 = util.img_as_ubyte(img2)
io.imsave(path, img2, plugin="pil", quality=100)
def histogram_equalization_and_save(file, img):
path = os.path.join(img_dir, histeq_dir, file)
img_eq = exposure.equalize_hist(img)
img2 = util.img_as_ubyte(img_eq)
io.imsave(path, img2, plugin="pil", quality=100)
def shear_image_and_save(file, img, shear, translation):
path = os.path.join(img_dir, shear_dir, file)
tform = transform.AffineTransform(shear=shear, translation=translation)
img_warped = transform.warp(img, tform)
img2 = util.img_as_ubyte(img_warped)
io.imsave(path, img2, plugin="pil", quality=100)
def save_original(file, img):
path = os.path.join(img_dir, orig_dir, file)
io.imsave(path, img, plugin="pil", quality=100)
count = 0
for entry in os.scandir(args.dir):
if entry.is_file() and entry.name.endswith(".jpg"):
img = io.imread(entry.name, plugin='pil')
if len(img.shape) == 3 and img.shape[2] >= 3:
print("({0}) : {1}".format(count, entry.name))
save_original(entry.name, img)
blur_image_and_save(entry.name, img, sigma=1.25)
compress_image_and_save(entry.name, img, quality=35)
crop_image_and_save(entry.name, img, pct_margin=0.10)
decimate_image_and_save(entry.name, img, (2, 2, 1))
noise_image_and_save(entry.name, img, gaussian_var=0.005)
occlude_image_and_save(entry.name, img, font_size=60)
rotate_image_and_save(entry.name, img, degrees=5.0)
scale_image_and_save(entry.name, img, downscale_dir, scale_factor=0.70)
scale_image_and_save(entry.name, img, upscale_dir, scale_factor=1.30)
hflip_image_and_save(entry.name, img)
vflip_image_and_save(entry.name, img)
gamma_correction_and_save(entry.name, img, darken_dir, gamma=2.2)
gamma_correction_and_save(entry.name, img, brighten_dir, gamma=0.4)
histogram_equalization_and_save(entry.name, img)
shear_image_and_save(entry.name, img, math.pi/8, (20, 10))
os.remove(entry.name)
count = count + 1
print("{0} files".format(count))
print("Done.")