-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathoperations_np.py
355 lines (250 loc) · 11.3 KB
/
operations_np.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
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
# from layers import batch_conv2d
from utils import *
import cv2
from scipy import ndimage
from matplotlib import colors
def apply_color_curve_adjustment(im, color_param):
L = color_param.shape[-2]
color_curve_sum = np.sum(color_param, axis=2, keepdims=True) + 1e-9
total_image = im * 0
for i in range(L):
color_curve_i = color_param[:, :, i, :]
total_image += np.clip(im - 1.0 * i / L, 0, 1.0 / L) * color_curve_i[:, None, :, :]
total_image *= L / color_curve_sum
return total_image
def apply_tone_curve_adjustment(im, tone_param):
L = tone_param.shape[-2]
tone_curve_sum = np.sum(tone_param, axis=2, keepdims=True) + 1e-9
total_image = im * 0
for i in range(L):
total_image += np.clip(im - 1.0 * i / L, 0, 1.0 / L) * tone_param[:, :, i, :]
total_image *= L / tone_curve_sum
return total_image
def apply_BnW(im, BnW_param):
luminance = rgb2lum(im)
BnW_im = lerp(im, luminance, BnW_param)
return BnW_im
def apply_saturation(im, saturation_param):
im = np.clip(im.squeeze(), 0, 1)
hsv = colors.rgb_to_hsv(im)
s = hsv[:, :, 1:2]
v = hsv[:, :, 2:3]
enhanced_s = s + (1 - s) * (0.5 - np.abs(0.5 - v)) * 0.8
hsv1 = np.concatenate([hsv[:, :, 0:1], enhanced_s, hsv[:, :, 2:]], axis=2)
full_color = colors.hsv_to_rgb(hsv1)
saturated_im = lerp(im, full_color, saturation_param)
return np.expand_dims(saturated_im, axis=0)
def apply_contrast(im, contrast_param):
luminance = np.clip(rgb2lum(im), 0, 1)
contrast_lum = -np.cos(np.pi * luminance) * 0.5 + 0.5
contrast_image = im / (luminance + 1e-6) * contrast_lum
return lerp(im, contrast_image, contrast_param)
def apply_WB(im, whiteBalanceparam):
rgb_means = np.mean(im, axis=(1, 2), keepdims=True) + 1e-9
balancing_vec = 0.5 / rgb_means
balancing_mat = np.tile(balancing_vec, [1, im.shape[1], im.shape[2], 1])
WB_im = im * balancing_mat
return lerp(im, WB_im, whiteBalanceparam)
def apply_exposure(im, exposureparam):
exposed_im = im * np.exp(exposureparam * np.log(2))
return exposed_im
def apply_affine(im, biasparam, slopeparam):
biasparam = biasparam[:, :, None, None]
biasparam = np.tile(biasparam, [1, im.shape[1], im.shape[2], im.shape[3]])
slopeparam = slopeparam[:, :, None, None]
slopeparam = np.tile(slopeparam, [1, im.shape[1], im.shape[2], im.shape[3]])
affine_im = slopeparam*im + biasparam
return affine_im
def apply_slope(im, slopeparam):
slopeparam = slopeparam[:, :, None, None]
slopeparam = np.tile(slopeparam, [1, im.shape[1], im.shape[2], im.shape[3]])
slope_im = slopeparam*im
return slope_im
def apply_blur(im, blurparam, ks=51):
blurred_im = cv2.GaussianBlur(im.squeeze(), (ks, ks), sigmaX=blurparam, sigmaY=blurparam,
borderType=cv2.BORDER_DEFAULT)
return blurred_im
# def apply_blur(im, blurparam, ks=51):
#
# im = np.clip(im, 0, 1)
# blur_vec = [np.exp((-(float(k) - ks)**2)/ (2 * blurparam**2 + 1e-7)) for k in range(ks * 2 + 1)]
# blur_vec = np.array(blur_vec)
# blur_vec = blur_vec/(np.sum(blur_vec, axis=-1, keepdims=True)+1e-7)
# blur_kernel = np.matmul(blur_vec, np.transpose(blur_vec))
# blur_kernel = np.stack([blur_kernel] * 3, axis=-1)
# blur_kernel = blur_kernel[None,:,:]
# blurred_im = ndimage.convolve(im, blur_kernel)/np.size(blur_kernel)
#
# return blurred_im
def apply_gamma(im, gammaparam):
gamma_im = im**gammaparam
return gamma_im
def apply_bias(im, biasparam):
bias_im = im + biasparam
return bias_im
def apply_sharpening_fg(im, sharpparam):
tf1 = (1 / 8) * np.array([[-1, 0, 1], [-2, 0, 2], [-1, 0, 1]])
tf2 = (1 / 8) * np.array([[-1, -2, -1], [0, 0, 0], [1, 2, 1]])
im1 = ndimage.convolve(im, np.expand_dims(np.stack([tf1]*3, axis=-1), axis=0))
im2 = ndimage.convolve(im, np.expand_dims(np.stack([tf2]*3, axis=-1), axis=0))
imedges = np.sqrt(im1**2 + im2**2)
sharp_im = im + sharpparam * imedges * im
return sharp_im
def apply_sharpening_bg(im, sharpparam):
tf1 = (1 / 8) * np.array([[-1, 0, 1], [-2, 0, 2], [-1, 0, 1]])
tf2 = (1 / 8) * np.array([[-1, -2, -1], [0, 0, 0], [1, 2, 1]])
im1 = ndimage.convolve(im, np.expand_dims(np.stack([tf1]*3, axis=-1), axis=0))
im2 = ndimage.convolve(im, np.expand_dims(np.stack([tf2]*3, axis=-1), axis=0))
imedges = np.sqrt(im1**2 + im2**2 + 1e-7)
sharp_im = im - sharpparam * imedges * im
return sharp_im
def apply_sharpening(im, sharpparam):
tf1 = (1 / 8) * np.array([[-1, 0, 1], [-2, 0, 2], [-1, 0, 1]])
tf2 = (1 / 8) * np.array([[-1, -2, -1], [0, 0, 0], [1, 2, 1]])
im1 = ndimage.convolve(im, np.expand_dims(np.stack([tf1]*3, axis=-1), axis=0))
im2 = ndimage.convolve(im, np.expand_dims(np.stack([tf2]*3, axis=-1), axis=0))
imedges = np.sqrt(im1**2 + im2**2)
sharp_im = im + sharpparam * imedges * im
return sharp_im
def apply_transformations_np(im, mask, params_fg, params_bg):
fg_im = apply_gamma(im, params_fg['gamma'])
bg_im = apply_gamma(im, params_bg['gamma'])
# Apply sharpening#
fg_im = apply_sharpening_fg(fg_im, params_fg['sharp'])
bg_im = apply_sharpening_bg(bg_im, params_bg['sharp'])
# Apply WB#
fg_im = apply_WB(fg_im, params_fg['WB'])
bg_im = apply_WB(bg_im, params_bg['WB'])
# Apply exposure #
fg_im = apply_exposure(fg_im, params_fg['exposure'])
bg_im = apply_exposure(bg_im, params_bg['exposure'])
# Apply contrast #
fg_im = apply_contrast(fg_im, params_fg['contrast'])
bg_im = apply_contrast(bg_im, params_bg['contrast'])
# Apply saturation #
fg_im = apply_saturation(fg_im, params_fg['saturation'])
bg_im = apply_saturation(bg_im, params_bg['saturation'])
# Apply tone curve adjustement#
fg_im = apply_tone_curve_adjustment(fg_im, params_fg['tone'])
bg_im = apply_tone_curve_adjustment(bg_im, params_bg['tone'])
# Apply color curve adjustement#
fg_im = apply_color_curve_adjustment(fg_im, params_fg['color'])
bg_im = apply_color_curve_adjustment(bg_im, params_bg['color'])
# Apply blur
bg_im = apply_blur(bg_im, params_bg['blur']+1e-9)
final_im = fg_im * mask + bg_im * (1 - mask)
final_im = np.clip(final_im, 0, 1)
# final_im = (final_im - 0.5) * 2
return final_im
def apply_transformations_np_gui(im, mask, params_fg, params_bg):
# Apply sharpening#
fg_im = apply_sharpening_fg(im, params_fg['sharp_fg'])
bg_im = apply_sharpening_bg(im, params_bg['sharp_bg'])
im_sharp = fg_im * mask + bg_im * (1 - mask)
# Apply exposure #
fg_im = apply_exposure(im_sharp, params_fg['exposure_fg'])
bg_im = apply_exposure(im_sharp, params_bg['exposure_bg'])
im_exp = fg_im * mask + bg_im * (1 - mask)
# Apply contrast #
fg_im = apply_contrast(im_exp, params_fg['contrast_fg'])
bg_im = apply_contrast(im_exp, params_bg['contrast_bg'])
im_cont = fg_im * mask + bg_im * (1 - mask)
# Apply tone curve adjustement#
fg_im = apply_tone_curve_adjustment(im_cont, params_fg['tone_fg'])
bg_im = apply_tone_curve_adjustment(im_cont, params_bg['tone_bg'])
im_tone = fg_im * mask + bg_im * (1 - mask)
# Apply color curve adjustement#
fg_im = apply_color_curve_adjustment(im_tone, params_fg['color_fg'])
bg_im = apply_color_curve_adjustment(im_tone, params_bg['color_bg'])
im_color = fg_im * mask + bg_im * (1 - mask)
# Apply blur
# bg_im = apply_blur(bg_im, params_bg['blur_bg']+1e-9)
final_im = np.clip(im_color, 0, 1)
# final_im = (final_im - 0.5) * 2
return final_im
def apply_transformations_np_nobias(im, mask, params_fg, params_bg):
fg_im = apply_gamma(im, params_fg['gamma_fg'])
bg_im = apply_gamma(im, params_bg['gamma_bg'])
# Apply sharpening#
fg_im = apply_sharpening_fg(fg_im, params_fg['sharp_fg'])
bg_im = apply_sharpening_bg(bg_im, params_bg['sharp_bg'])
# Apply bias#
# fg_im = apply_bias(fg_im, params_fg['bias_fg'])
# bg_im = apply_bias(bg_im, params_bg['bias_bg'])
# Apply WB#
fg_im = apply_WB(fg_im, params_fg['WB_fg'])
bg_im = apply_WB(bg_im, params_bg['WB_bg'])
# Apply exposure #
fg_im = apply_exposure(fg_im, params_fg['exposure_fg'])
bg_im = apply_exposure(bg_im, params_bg['exposure_bg'])
# Apply contrast #
fg_im = apply_contrast(fg_im, params_fg['contrast_fg'])
bg_im = apply_contrast(bg_im, params_bg['contrast_bg'])
# Apply saturation #
fg_im = apply_saturation(fg_im, params_fg['saturation_fg'])
bg_im = apply_saturation(bg_im, params_bg['saturation_bg'])
# Apply BnW adjustement#
fg_im = apply_BnW(fg_im, params_fg['BnW_fg'])
bg_im = apply_BnW(bg_im, params_bg['BnW_bg'])
# Apply tone curve adjustement#
fg_im = apply_tone_curve_adjustment(fg_im, params_fg['toneAdjustement_fg'])
bg_im = apply_tone_curve_adjustment(bg_im, params_bg['toneAdjustement_bg'])
# Apply color curve adjustement#
fg_im = apply_color_curve_adjustment(fg_im, params_fg['colorAdjustement_fg'])
bg_im = apply_color_curve_adjustment(bg_im, params_bg['colorAdjustement_bg'])
# Apply blur
bg_im = apply_blur(bg_im, params_bg['blur_bg']+1e-9)
final_im = fg_im * mask + bg_im * (1 - mask)
final_im = np.clip(final_im, 0, 1)
# final_im = (final_im - 0.5) * 2
return final_im
def apply_transformations_np_decrease(im, mask, params_fg, params_bg):
fg_im = apply_gamma(im, params_fg['gamma_fg'])
bg_im = apply_gamma(im, params_bg['gamma_bg'])
# Apply sharpening#
fg_im = apply_sharpening_fg(fg_im, params_fg['sharp_fg'])
bg_im = apply_sharpening_bg(bg_im, params_bg['sharp_bg'])
# Apply bias#
# fg_im = apply_bias(fg_im, params_fg['bias_fg'])
# bg_im = apply_bias(bg_im, params_bg['bias_bg'])
# Apply WB#
fg_im = apply_WB(fg_im, params_fg['WB_fg'])
bg_im = apply_WB(bg_im, params_bg['WB_bg'])
# Apply exposure #
fg_im = apply_exposure(fg_im, params_fg['exposure_fg'])
bg_im = apply_exposure(bg_im, params_bg['exposure_bg'])
# Apply contrast #
fg_im = apply_contrast(fg_im, params_fg['contrast_fg'])
bg_im = apply_contrast(bg_im, params_bg['contrast_bg'])
# Apply saturation #
fg_im = apply_saturation(fg_im, params_fg['saturation_fg'])
bg_im = apply_saturation(bg_im, params_bg['saturation_bg'])
# Apply BnW adjustement#
fg_im = apply_BnW(fg_im, params_fg['BnW_fg'])
bg_im = apply_BnW(bg_im, params_bg['BnW_bg'])
# Apply tone curve adjustement#
fg_im = apply_tone_curve_adjustment(fg_im, params_fg['toneAdjustement_fg'])
bg_im = apply_tone_curve_adjustment(bg_im, params_bg['toneAdjustement_bg'])
# Apply color curve adjustement#
fg_im = apply_color_curve_adjustment(fg_im, params_fg['colorAdjustement_fg'])
bg_im = apply_color_curve_adjustment(bg_im, params_bg['colorAdjustement_bg'])
# Apply blur
# bg_im = apply_blur(bg_im, params_bg['blur_bg']+1e-9)
final_im = fg_im * mask + bg_im * (1 - mask)
final_im = np.clip(final_im, 0, 1)
# final_im = (final_im - 0.5) * 2
return final_im
def get_margins(m):
tmp_x = np.nonzero(np.sum(m, axis=0))
tmp_y = np.nonzero(np.sum(m, axis=1))
tmp_x = tmp_x[0]
tmp_y = tmp_y[0]
try:
margins_x = [np.min(tmp_x), np.max(tmp_x)]
margins_y = [np.min(tmp_y), np.max(tmp_y)]
cont = False
except:
margins_x = 0
margins_y = 0
cont = True
return margins_x, margins_y, cont