-
Notifications
You must be signed in to change notification settings - Fork 34
/
rawdata_processing.py
196 lines (166 loc) · 8.34 KB
/
rawdata_processing.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
import numpy as np
import matplotlib.pyplot as plt
import cv2
from glob import glob
import rawpy
import os
import exifread
def demosaic(raw_array, exif_dict=None):
if str(exif_dict["Image Make"]) == "HUAWEI":
rgb = np.stack([raw_array[1::2, 1::2], (raw_array[0::2, 1::2] + raw_array[1::2, 0::2]) / 2,
raw_array[0::2, 0::2]], axis=2)
else:
rgb = np.stack([raw_array[0::2, 0::2], (raw_array[0::2, 1::2] + raw_array[1::2, 0::2]) / 2,
raw_array[1::2, 1::2]], axis=2)
neutral_wb =tag2matrix(str(exif_dict["EXIF:AsShotNeutral"]))
rgb = np.concatenate([rgb[...,0:1].clip(0,neutral_wb[0,0]),
rgb[...,1:2].clip(0,neutral_wb[1,0]),
rgb[...,2:3].clip(0,neutral_wb[2,0])], axis=2)
return rgb#.clip(0,1.0)
def Linearization(raw_bayer, exif_dict):
black_level = exif_dict["BlackLevel"]
white_level = exif_dict["WhiteLevel"]
raw_bayer = raw_bayer.astype(np.float32)
raw_bayer = (raw_bayer - black_level) / (white_level - black_level)
raw_linear = raw_bayer
return raw_linear
def gamma_correction(rgb, gamma=2.2):
return np.power(rgb, 1 / gamma)
def get_matrix(m1, m2, tp1, tp2, tp):
if (tp < tp1):
m = m1
elif (tp > tp2):
m = m2
else:
g = (1/ float(tp) - 1 / float(tp2)) / (1 / float(tp1) - 1 / float(tp2))
m = g * m1 + (1-g) * m2
return m
def WhiteBalance_ColorCalibration(rgb_demosaic, exif_dict):
d50tosrgb = np.reshape(np.array([3.1338561, -1.6168667, -0.4906146,
-0.9787684, 1.9161415, 0.0334540,
0.0719453, -0.2289914, 1.4052427]), (3,3))
d50toprophotorgb = np.reshape(np.array([1.3459433, -0.2556075, -0.0511118,
-0.5445989, 1.5081673, 0.0205351,
0, 0, 1.2118128]), (3,3))
height, width, channels = rgb_demosaic.shape
forward_matrix1 = tag2matrix(str(exif_dict["EXIF:ForwardMatrix1"]))
forward_matrix2 = tag2matrix(str(exif_dict["EXIF:ForwardMatrix2"]))
color_matrix1 = tag2matrix(str(exif_dict["EXIF:ColorMatrix1"]))
color_matrix2 = tag2matrix(str(exif_dict["EXIF:ColorMatrix2"]))
camera_calibration1 = tag2matrix(str(exif_dict["EXIF:CameraCalibration1"]))
camera_calibration2 = tag2matrix(str(exif_dict["EXIF:CameraCalibration2"]))
neutral_wb =tag2matrix(str(exif_dict["EXIF:AsShotNeutral"]))
analog_balance = np.diag(np.asarray([float(i) for i in exif_dict["EXIF:AnalogBalance"].values])) #np.diag([1, 1, 1])
rgb_demosaic = np.concatenate([rgb_demosaic[...,0:1].clip(0,neutral_wb[0,0]),
rgb_demosaic[...,1:2].clip(0,neutral_wb[1,0]),
rgb_demosaic[...,2:3].clip(0,neutral_wb[2,0])], axis=2)
# Standard light A
temparature1 = 2850
# D65
temparature2 = 6500
if (exif_dict["EXIF:Make"] == "NIKON CORPORATION"):
image_temparatue = 4000#exif_dict["MakerNotes:ColorTemperatureAuto"]
elif (exif_dict["EXIF:Make"] == "HUAWEI"):
# hack for HUAWEI
image_temparatue = 4000
forward_matrix = get_matrix(forward_matrix1, forward_matrix2, temparature1, temparature2, image_temparatue)
camera_calibration = get_matrix(camera_calibration1, camera_calibration2, temparature1, temparature2, image_temparatue)
rgb_reshaped = np.reshape(np.transpose(rgb_demosaic, (2,0,1)),(3,-1))
prophotorgbtod50 = np.linalg.inv(d50toprophotorgb)
ref_neutral = np.matmul(np.linalg.inv(np.matmul(analog_balance,camera_calibration)), neutral_wb)
d = np.linalg.inv(np.diag([ref_neutral[0,0], ref_neutral[1,0], ref_neutral[2,0]]))
camera2d50 = np.matmul(np.matmul(forward_matrix, d),
np.linalg.inv(np.matmul(analog_balance, camera_calibration)))
camera2srgb = np.matmul(d50tosrgb, camera2d50)
rgb_srgb = np.matmul(camera2srgb, rgb_reshaped)
orgshape_rgb_srgb = np.reshape(np.transpose(rgb_srgb, (1, 0)),(height, width,3))
return orgshape_rgb_srgb.clip(0, 1)
# Step 7: Applying the hue / saturation / value mapping
# Step 9: apply color mapping - no for HUAWEI
# Step 10:Tone curve
# Fitting curve
# If the tone curve is not found, we use a default tone curve
# Step 11: Convert to srgb
# Step 12: Gamma Correction
def tag2matrix(info):
value_list = info[1:-1].split(", ")
value_array = []
for value in value_list:
if "/" in value:
den, num = value.split("/")
value_array.append(float(den) / float(num))
else:
value_array.append(float(value))
if len(value_list) == 9:
value_mat = np.array(value_array).reshape([3,3])
else:
value_mat = np.array(value_array).reshape([3,1])
return value_mat
def prepare_exifdict(raw_path):
# Open image file for reading (binary mode)
f = open(raw_path, 'rb')
# Return Exif tags
exif_dict = exifread.process_file(f, details=True)
exif_dict["EXIF:Make"] = str(exif_dict["Image Make"])
if str(exif_dict["Image Make"]) == "NIKON CORPORATION":
exif_dict["BlackLevel"] = 1008.0
exif_dict["WhiteLevel"] = 16384.0
elif str(exif_dict["Image Make"]) == "HUAWEI":
exif_dict["BlackLevel"] = 256.0
exif_dict["WhiteLevel"] = 4096.0
exif_dict["EXIF:ColorMatrix1"] = exif_dict["Image Tag 0xC621"]
exif_dict["EXIF:ColorMatrix2"] = exif_dict["Image Tag 0xC622"]
exif_dict["EXIF:CameraCalibration1"] = exif_dict["Image Tag 0xC623"]
exif_dict["EXIF:CameraCalibration2"] = exif_dict["Image Tag 0xC624"]
exif_dict["EXIF:ForwardMatrix1"] = exif_dict["Image Tag 0xC714"]
exif_dict["EXIF:ForwardMatrix2"] = exif_dict["Image Tag 0xC715"]
exif_dict["EXIF:AsShotNeutral"] = exif_dict["Image Tag 0xC628"]
exif_dict["EXIF:AnalogBalance"] = exif_dict["EXIF:AsShotNeutral"]
return exif_dict
def prepare_rawlinear(raw_path, norm = False):
exif_dict = prepare_exifdict(raw_path)
raw_bayer = rawpy.imread(raw_path).raw_image_visible.copy()
raw_linear = Linearization(raw_bayer.copy(), exif_dict)
if norm == True:
raw_nonexposure = raw_linear.copy()
raw_nonexposure[raw_nonexposure>0.75] = 0
raw_linear = raw_linear * (1.0 / raw_nonexposure.max())
rgb_demosaic = demosaic(raw_linear, exif_dict)
return rgb_demosaic
def process_raw_from_raw_linear(rgb_demosaic, exif_dict):
orgshape_rgb_srgb = WhiteBalance_ColorCalibration(rgb_demosaic, exif_dict)
rgb_gamma = gamma_correction(orgshape_rgb_srgb)
return rgb_gamma
def obtain_rgb_flashonly(path_A, path_B, raw_camera="Huawei", norm=False):
input_raw1 = path_A
input_raw2 = path_B
if raw_camera == "Huawei":
raw_flash_demosaic = prepare_rawlinear(input_raw1)
raw_ambient_demosaic = prepare_rawlinear(input_raw2)
raw_pureflash = (raw_flash_demosaic - raw_ambient_demosaic).clip(0, 1)
if norm == True:
raw_nonexposure = raw_flash_demosaic.copy()
raw_nonexposure[raw_nonexposure>0.75] = 0
ratio = 1.0 / raw_nonexposure.max()
raw_pureflash = raw_pureflash * ratio
raw_flash_demosaic = raw_flash_demosaic * ratio
raw_ambient_demosaic = raw_ambient_demosaic * ratio
exif_dict = prepare_exifdict(input_raw2)
rgb_A_minus_B = process_raw_from_raw_linear(raw_pureflash, exif_dict)
rgb_A = process_raw_from_raw_linear(raw_flash_demosaic, exif_dict)
rgb_B = process_raw_from_raw_linear(raw_ambient_demosaic, exif_dict)
else:
black_level = {"Huawei":256, "Nikon":1024}
with rawpy.imread(input_raw1) as bayer1:
rgb_A=bayer1.postprocess()
bayer2 = rawpy.imread(input_raw2)
rgb_B=bayer2.postprocess()
bayer1.raw_image_visible[:] = bayer1.raw_image_visible[:] - bayer2.raw_image_visible[:] + black_level[raw_camera]
rgb_A_minus_B = bayer1.postprocess()
return rgb_A, rgb_B, rgb_A_minus_B
raw_ambient_path = "flash.dng"
raw_flash_path = "ambient.dng"
rgb_flash, rgb_ambient, rgb_flashonly = obtain_rgb_flashonly(raw_flash_path, raw_ambient_path)
cv2.imwrite("rgb_ambient.jpg", rgb_ambient[::10,::10,::-1] * 255.)
cv2.imwrite("rgb_flash.jpg", rgb_flash[::10,::10,::-1]* 255.)
cv2.imwrite("rgb_flashonly.jpg", rgb_flashonly[::10,::10,::-1]* 255.)