-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathCC_IQA.py
380 lines (341 loc) · 15 KB
/
CC_IQA.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
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
from collections import Counter
import cv2
import os
import numpy as np
import matplotlib.pyplot as plt
from sklearn.cluster import KMeans
from ColorAnalysis import cba # cba: ColorBlock Analysis
from colormath.color_conversions import convert_color
from colormath.color_objects import sRGBColor, LabColor
from colormath.color_diff import delta_e_cie2000
#### The data from https://www.xrite.com/service-support/new_color_specifications_for_colorchecker_sg_and_classic_charts
CIE_lab_24 = [[37.54, 14.37, 14.92], [64.66, 19.27, 17.5], [49.32, -3.82, -22.54], [43.46, -12.74, 22.72], [54.94, 9.61, -24.79], [70.48, -32.26, -0.37],
[62.73, 35.83, 56.5], [39.43, 10.75, -45.17], [50.57, 48.64, 16.67], [30.1, 22.54, -20.87], [71.77, -24.13, 58.19], [71.51, 18.24, 67.37],
[28.37, 15.42, -49.8], [54.38, -39.72, 32.27], [42.43, 51.05, 28.62], [81.8, 2.67, 80.41], [50.63, 51.28, -14.12], [49.57, -29.71, -28.32],
[95.19, -1.03, 2.93], [81.29, -0.57, 0.44], [66.89, -0.75, -0.06], [50.76, -0.13, 0.14], [35.63, -0.46, -0.48], [20.64, 0.07, -0.46]]
def rgb2lab(rgb):
r = rgb[0] / 255.0 # rgb range: 0 ~ 1
g = rgb[1] / 255.0
b = rgb[2] / 255.0
# gamma 2.2
if r > 0.04045:
r = pow((r + 0.055) / 1.055, 2.4)
else:
r = r / 12.92
if g > 0.04045:
g = pow((g + 0.055) / 1.055, 2.4)
else:
g = g / 12.92
if b > 0.04045:
b = pow((b + 0.055) / 1.055, 2.4)
else:
b = b / 12.92
# sRGB
X = r * 0.436052025 + g * 0.385081593 + b * 0.143087414
Y = r * 0.222491598 + g * 0.716886060 + b * 0.060621486
Z = r * 0.013929122 + g * 0.097097002 + b * 0.714185470
# XYZ range: 0~100
X = X * 100.000
Y = Y * 100.000
Z = Z * 100.000
# Reference White Point
ref_X = 96.4221
ref_Y = 100.000
ref_Z = 82.5211
X = X / ref_X
Y = Y / ref_Y
Z = Z / ref_Z
# Lab
if X > 0.008856:
X = pow(X, 1 / 3.000)
else:
X = (7.787 * X) + (16 / 116.000)
if Y > 0.008856:
Y = pow(Y, 1 / 3.000)
else:
Y = (7.787 * Y) + (16 / 116.000)
if Z > 0.008856:
Z = pow(Z, 1 / 3.000)
else:
Z = (7.787 * Z) + (16 / 116.000)
Lab_L = round((116.000 * Y) - 16.000, 2)
Lab_a = round(500.000 * (X - Y), 2)
Lab_b = round(200.000 * (Y - Z), 2)
return Lab_L, Lab_a, Lab_b
def color_difference(measure, ref):
L = abs(measure[0] - ref[0])
a = abs(measure[1] - ref[1])
b = abs(measure[2] - ref[2])
C = (a ** 2 + b ** 2) ** 0.5
E = (a ** 2 + b ** 2 + L ** 2) ** 0.5
return C, E
def draw_rect(img, lt, rb):
point_color = (0, 255, 0) # BGR
thickness = 2
lineType = 4
result = cv2.rectangle(img, lt, rb, point_color, thickness, lineType)
return result
def load_std_ref_LAB(c_type):
global REF_LAB
REF_LAB = []
if c_type == "24":
REF_LAB = CIE_lab_24
else:
assert True, "type error"
def get_delta_e(color1, color2):
#色彩空間轉換(BGR to LAB)
def rgb2lab_cie(rgb):
#print(f"Input RGB: {rgb}")
return convert_color(sRGBColor(rgb[0], rgb[1], rgb[2]), LabColor)
# 計算CIEDE2000的色差
delta_e_2000 = delta_e_cie2000(rgb2lab_cie(color1), rgb2lab_cie(color2))
return delta_e_2000
def remove_outliers(rgb_array, threshold):
rgb_np = np.array(rgb_array)
#print("rgb_np[0]")
#print(rgb_np[:,:,0])
#print("rgb_np[1]")
#print(rgb_np[:,:,1])
#print("rgb_np[2]")
#print(rgb_np[:,:,2])
# 找到每個通道的最小值和最大值
min_vals_r = np.min(rgb_np[:,:,0])
min_vals_g = np.min(rgb_np[:,:,1])
min_vals_b = np.min(rgb_np[:,:,2])
max_vals_r = np.max(rgb_np[:,:,0])
max_vals_g = np.max(rgb_np[:,:,1])
max_vals_b = np.max(rgb_np[:,:,2])
# print(f"min_vals_r:{min_vals_r}")
# print(f"min_vals_g:{min_vals_g}")
# print(f"min_vals_b:{min_vals_b}")
# print(f"max_vals_r:{max_vals_r}")
# print(f"max_vals_g:{max_vals_g}")
# print(f"max_vals_b:{max_vals_b}")
# 判斷是否為極值
is_outlier_r = np.logical_or(rgb_np[:,:,0] < min_vals_r + threshold, rgb_np[:,:,0] > max_vals_r - threshold)
#print("is_outlier_r:")
#print(is_outlier_r)
is_outlier_g = np.logical_or(rgb_np[:,:,1] < min_vals_g + threshold, rgb_np[:,:,1] > max_vals_g - threshold)
#print("is_outlier_g:")
#print(is_outlier_g)
is_outlier_b = np.logical_or(rgb_np[:,:,2] < min_vals_b + threshold, rgb_np[:,:,2] > max_vals_b - threshold)
#print("is_outlier_b:")
#print(is_outlier_b)
# 組合三個通道的極值判斷結果
is_outlier = np.logical_or.reduce([is_outlier_r, is_outlier_g, is_outlier_b])
# 返回篩選後的 RGB 值陣列
filtered_rgb = rgb_np[~is_outlier]
#print("filtered_rgb:")
#print(filtered_rgb)
return filtered_rgb
def cc_task(cc_img, scale=0.5):
##### The data from https://www.xrite.com/service-support/new_color_specifications_for_colorchecker_sg_and_classic_charts
load_std_ref_LAB("24")
h, w = cc_img.shape[0], cc_img.shape[1]
# 色板 RGB 值的列表
# 由右到左，由上到下顯示
rgb_list = [[51, 51, 51], [102, 102, 102], [153, 153, 153], [204, 204, 204], [255, 255, 255],
[0, 0, 51], [0, 0, 102], [0, 0, 153], [0, 0, 204], [0, 0, 255],
[0, 128, 255], [128, 255, 0], [255, 255, 0], [255, 190, 0], [255, 0, 0],
[0, 51, 0], [0, 102, 0], [0, 153, 0], [0, 204, 0], [0, 255, 0],
[128, 74, 0], [131, 0, 181], [255, 0, 255], [0, 255, 255], [0, 0, 0]]
h_block = (h / 7)
w_block = (w / 7)
center_points = []
real_center_points = []
for y in range(1, 7):
for x in range(1, 7):
center_points.append([x * w_block, y * h_block])
for i in range(6):
for j in range(6):
if i == 0 or j == 0:
continue
real_center_points.append(center_points[i * 6 + j])
center_points = real_center_points[0:24]
#print("center_points", center_points)
img_rect = cc_img
img_rect_ = img_rect.copy()
count = 0
mean_C = 0
mean_E = 0
center_rgb = []
center_rgb_clean = []
# rgb跟rgbc的差異
rgb_rgbc_cmp = []
# rgb跟原色板的差異
rgb_std_cmp = []
# rgbc跟原色板的差異
rgbc_std_cmp = []
for c in center_points:
# 解釋變數
# c: 中心點座標
# 中心點座標是指每個色塊的中心點
# 每個色塊的中心點是怎麼找的呢?
# 以 24 格的 ColorChecker 為例
# 24 格的 ColorChecker 是 4 * 6 的矩陣
# 每個色塊的寬度是整張圖片的 1/6
# 每個色塊的高度是整張圖片的 1/4
# 所以每個色塊的中心點座標是
# (1/6 * 1/2, 1/4 * 1/2) ~ (5/6 * 1/2, 3/4 * 1/2)
# c_w: 中心點寬度
# c_h: 中心點高度
# lt: 左上角座標
# rb: 右下角座標
# cc_block: 中心點區域
c_w = c[0] - w_block / 2
c_h = c[1] - h_block / 2
lt = (int(c_w - (w_block / 2) * scale), int(c_h - (h_block / 2) * scale))
rb = (int(c_w + (w_block / 2) * scale), int(c_h + (h_block / 2) * scale))
img_rect_ = draw_rect(img_rect_, lt, rb)
# print("lt, rb", lt, rb)
cc_block = img_rect[lt[1]: rb[1], lt[0]: rb[0]]
# print("cc_block", cc_block)
cc_block_clean = cc_block.copy()
cc_block_clean = remove_outliers(cc_block_clean, 1)
kmeans_ori = KMeans(n_clusters=5, random_state=0, n_init='auto').fit(cc_block.reshape(-1, 3))
kmeans_ori_clean = KMeans(n_clusters=5, random_state=0, n_init='auto').fit(remove_outliers(cc_block, 1).reshape(-1, 3))
kmeans_ori_label = kmeans_ori.labels_
kmeans_ori_label_clean = kmeans_ori_clean.labels_
cluster_sizes_ori = Counter(kmeans_ori_label)
cluster_sizes_ori_clean = Counter(kmeans_ori_label_clean)
largest_cluster_ori = max(cluster_sizes_ori, key=cluster_sizes_ori.get)
largest_cluster_ori_clean = max(cluster_sizes_ori_clean, key=cluster_sizes_ori_clean.get)
representative_color_ori = kmeans_ori.cluster_centers_[largest_cluster_ori]
representative_color_ori_clean = kmeans_ori_clean.cluster_centers_[largest_cluster_ori_clean]
R_mean = np.mean(cc_block[:, :, 0])
G_mean = np.mean(cc_block[:, :, 1])
B_mean = np.mean(cc_block[:, :, 2])
R_mean_clean = np.mean(cc_block_clean[:, 0])
G_mean_clean = np.mean(cc_block_clean[:, 1])
B_mean_clean = np.mean(cc_block_clean[:, 2])
# RGB_m: 中心點 RGB 值
RGB_m = [R_mean, G_mean, B_mean]
RGB_m_clean = [R_mean_clean, G_mean_clean, B_mean_clean]
center_rgb.append(RGB_m)
#[[r,g,b]*24個]
center_rgb_clean.append(RGB_m_clean)
rgb_rgbc_cmp.append(get_delta_e(representative_color_ori, representative_color_ori_clean))
rgb_std_cmp.append(get_delta_e(representative_color_ori, rgb_list[count]))
rgbc_std_cmp.append(get_delta_e(representative_color_ori_clean, rgb_list[count]))
# LAB_m: 中心點 LAB 值
LAB_m = rgb2lab(RGB_m)
#print(f"count: {count}")
#print("LAB_m: ", LAB_m, "REF_LAB: ", REF_LAB[count])
C, E = color_difference(LAB_m, REF_LAB[count])
mean_C += C
mean_E += E
count += 1
#------------------cie_2000------------------
#------- 色差比較圖 -------#
center_rgb_temp = center_rgb.copy()
center_rgb_temp = np.append(center_rgb_temp, [[0,0,0]], axis=0)
# 改成5x5x3陣列
center_rgb_2d = np.array(center_rgb_temp).reshape(5,5,3)
# print(center_rgb_2d)
rgb_list_float = np.array(rgb_list).reshape(5,5,3)
# print(rgb_list_float)
delta_e = cba.compare_colorboard(rgb_list_float, center_rgb_2d)
plt.savefig(os.path.join('res', "delta_e.png")) # 儲存色差比較圖
# fio.save_image_file('delta_e', 'res') # 如果要把所有生成的色差比較圖都留下，就用這個
#------- 色差直方圖 -------#
delta_e = np.array(delta_e)
delta_e_mean = np.mean(delta_e[0:24]) # 去掉最後一個標記色塊
# 繪製直方圖
x = np.arange(0, 25)
labels = [f"({i//5},{i%5})" for i in range(25)]
fig, ax = plt.subplots(figsize=(14, 6))
ax.bar(x, delta_e.reshape(25), width=0.4, label='delta_e: restored_model')
# 繪製平均值
ax.axhline(delta_e_mean, color='r', linestyle='--', label='delta_e mean')
ax.text(24.6, delta_e_mean, f"{delta_e_mean:.2f}", ha='right', va='bottom')
# 設定圖表屬性
ax.set_xticks(x+0.4/2)
ax.set_xticklabels(labels)
ax.legend()
ax.set_title('Histogram of delta_e')
ax.set_xlabel('(i,j)')
ax.set_ylabel('Value')
plt.savefig(os.path.join('res', 'Histogram of delta_e')) # 儲存色差直方圖
# fio.save_image_file('Histogram of delta_e', result_dir) # 如果要把所有生成的色差直方圖都留下，就用這個
#------------------cie_2000------------------
# 由於中心點的 RGB 值會受到環境光影響
# 所以我們需要將 RGB 值濾掉離群值
# 這裡我們使用的方法是
# 將 RGB 值轉換成 LAB 值
# 再將 LAB 值轉換成 XYZ 值
# 再將 XYZ 值轉換成 LAB 值
# 再將 LAB 值轉換成 RGB 值
# 這樣就可以將 RGB 值濾掉離群值
# 這裡我們使用的是 OpenCV 的函式
# 這裡的函式會將 RGB 值轉換成 XYZ 值
# 再將 XYZ 值轉換成 LAB 值
# 再將 LAB 值轉換成 RGB 值
center_rgb = np.array(center_rgb)
center_rgb_clean = np.array(center_rgb_clean)
rgb_list = np.array(rgb_list)
_, axs = plt.subplots(1, 3, figsize=(24, 8))
#------------------k-means------------------
#------------------每一點的rgb值與色差------------------
axs[0].set_title('k-means')
axs[1].set_title('k-means_clean')
axs[2].set_title('original')
for i in range(5):
for j in range(5):
if i == 4 and j == 4:
break
x = j*4
y = (5-i)*4
# 由於 k-means 分群的結果是隨機的
# 所以要將分群結果和色板的顏色進行比較
axs[0].add_patch(plt.Rectangle((x, y), 4, 4, color=np.array(center_rgb[i * 5 + j]) / 255))
axs[1].add_patch(plt.Rectangle((x, y), 4, 4, color=np.array(center_rgb_clean[i * 5 + j]) / 255))
axs[2].add_patch(plt.Rectangle((x, y), 4, 4, color=np.array(rgb_list[i * 5 + j]) / 255))
# 在方框中心位置添加 RGB 值文字
rgb_text = f"({round(center_rgb[i * 5 + j][0])},{round(center_rgb[i * 5 + j][1])},{round(center_rgb[i * 5 + j][2])})"
axs[0].text(x + 2, y + 2, rgb_text, color='black', fontsize=13, ha='center',
va='center', bbox=dict(facecolor='white', edgecolor='black', boxstyle='round'))
rgb_clean_text = f"({round(center_rgb_clean[i * 5 + j][0])},{round(center_rgb_clean[i * 5 + j][1])},{round(center_rgb_clean[i * 5 + j][2])})"
axs[1].text(x + 2, y + 2, rgb_clean_text, color='black', fontsize=13, ha='center',
va='center', bbox=dict(facecolor='white', edgecolor='black', boxstyle='round'))
rgb_text = f"({round(rgb_list[i * 5 + j][0])},{round(rgb_list[i * 5 + j][1])},{round(rgb_list[i * 5 + j][2])})"
axs[2].text(x + 2, y + 2, rgb_text, color='black', fontsize=13, ha='center',
va='center', bbox=dict(facecolor='white', edgecolor='black', boxstyle='round'))
axs[0].axis('scaled')
axs[1].axis('scaled')
axs[2].axis('scaled')
plt.tight_layout()
plt.savefig('res/colorblock.png', dpi=300, bbox_inches='tight')
# 將 Delta E 值和顏色對應的索引轉換為 NumPy 陣列
delta_e_rgb_rgbc = np.array(rgb_rgbc_cmp)
delta_e_rgb_std = np.array(rgb_std_cmp)
delta_e_rgbc_std = np.array(rgbc_std_cmp)
color_indices = np.arange(len(rgb_list) - 1)
plt.figure(figsize=(12, 6))
plt.subplot(131)
plt.bar(color_indices, delta_e_rgb_rgbc)
# 設定圖表標籤和標題
plt.xlabel('Color Index')
plt.ylabel('Delta E')
plt.title('Colors and Remove outlier Color')
plt.subplot(132)
plt.bar(color_indices, delta_e_rgb_std)
# 設定圖表標籤和標題
plt.xlabel('Color Index')
plt.ylabel('Delta E')
plt.title('Colors and Standard Color')
plt.subplot(133)
plt.bar(color_indices, delta_e_rgbc_std)
# 設定圖表標籤和標題
plt.xlabel('Color Index')
plt.ylabel('Delta E')
plt.title('Remove outlier Colors and Standard Color')
plt.savefig('res/k-means.png', dpi=300, bbox_inches='tight')
#------------------每一點的rgb值與色差------------------
mean_C /= 24
mean_E /= 24
# print("C:{} | E:{} ".format(mean_C, mean_E))
#print(mean_C, mean_E)
return {"mean_C" : mean_C, "mean_E" : mean_E, "img_rect_" : img_rect_,
"center_rgb" : center_rgb, "center_rgb_clean" : center_rgb_clean, "rgb_list" : rgb_list,
"delta_e_rgb_rgbc" : delta_e_rgb_rgbc, "delta_e_rgb_std" : delta_e_rgb_std, "delta_e_rgbc_std" : delta_e_rgbc_std,
"color_indices" : color_indices}