Coralie Cauwerts, Sophie Jost, and Bertrand Deroisy, "Calibration of high dynamic range images for applied color and lighting research," J. Opt. Soc. Am. A 36, C130-C142 (2019)
The main objective of this work is to determine the most appropriate high dynamic range picture calibration procedure for measuring luminance and color, in the context of architecture. Three professional digital single-lens reflex cameras fitted with fisheye lenses were tested. Photometric and colorimetric accuracy was assessed in comparison with spectroradiometer measurements of 57 color samples (Macbeth chart and additional Munsell samples). The results demonstrate that using a calibration model specific to the camera–lens association rather than the standardized RGB to XYZ color transform matrix is necessary to achieve an acceptable colorimetric accuracy. Moreover, using a specific color transform matrix reduces the error in luminance, especially for colorful elements. The study also shows the opportunity to share the same color transform matrix between similar photographic materials (same brand and model). Finally, among the three tested devices, one camera–lens association has higher performance and produces better quality QuickTime Virtual Reality panoramas.
You do not have subscription access to this journal. Cited by links are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
You do not have subscription access to this journal. Figure files are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
You do not have subscription access to this journal. Article tables are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
You do not have subscription access to this journal. Equations are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
Fisheye lens.
Not specified.
CTM determined using an ultra-cool daylight source (CCT not specified).
CTM determined using a cool incandescent source (5050K) + 3CF.
Ultra-cool daylight + real room lighting.
HDR assembly part of Debevec and Malik’s method [10].
CTM determined at 5500K.
Tables (9)
Table 1.
Specifications of the Three Tested Cameras and Picture Quality (Large/Fine Setting)
Camera
Setting
Canon EOS 40D
Canon EOS 50D
Canon EOS 5D Mark II
Sensor size
Sensor type
CMOS APS-C
CMOS APS-C
CMOS 35 mm
Crop factor
1.6
1.6
1
Lens mount type
EF, EF-S
EF, EF-S
EF
Pixels
Definition
File size (.jpg)
File size (.cr2)
Table 2.
Camera Settings
Parameters
Mode
White balance
Daylight ( in the Canon system)
Sensitivity
ISO 100
Focusing setting
Manual ()
Picture quality
raw
Exposure mode
Manual
Table 3.
Capture Settings
Scene
Device
ISO
Av
Tv
Training
40Dfe45
100
f/10
1.6, 0.8, ½, 1/5, 1/10, 1/20, 1/40
Training
50D fe45
100
f/10
1.6, 0.8, ½, 1/5, 1/10, 1/20, 1/40, 1/80
Training
5Dfe8
100
f/16
8, 4, 2, 1, 0.5, 1/4, 1/8, 1/15
Test
40Dfe45
100
f/10
½, 1/5, 1/10, 1/20, 1/40, 1/80, 1/160
Test
50D fe45
100
f/10
½, 1/5, 1/10, 1/20, 1/40, 1/80, 1/160
Test
5Dfe8
100
f/16
2,1,1/2,/1/4,1/8, 1/15, 1/30, 1/60
Table 4.
Calibration Factors
Device
40Dfe45
50Dfe45
5Dfe8
CTM
Re-scaling
CF_X
CF_Y
CF_Z
CF_X
CF_Y
CF_Z
CF_X
CF_Y
CF_Z
sRGB
1CF
-
1.53
-
-
1.68
-
-
1.73
-
3CF
1.55
1.53
1.61
1.63
1.68
1.69
1.70
1.73
1.86
minXYZ
1CF
-
0.96
-
-
0.96
-
-
0.98
-
3CF
0.99
0.96
0.96
1.00
0.96
0.95
0.99
0.98
0.99
Table 5.
Mean (MAPE_lum) and Maximum Absolute Percentage Error of Luminance, by Device, for 57 Samples, with Test Source
Device
40Dfe45
50Dfe45
5Dfe8
Calibration Method
MAPE
maximum
MAPE
maximum
MAPE
maximum
sRGB (1CF/3CF)
5.5%
24.2% (#15-red)
4.2%
21.4% (#15-red)
3.6%
19.0% (#15-red)
minXYZ (1CF/3CF)
4.1%
21.2% (5PB 2.5/4)
3.6%
10.7% (5RP 2.5/4)
2.7%
8.7% (#13-blue)
Table 6.
Analysis of Variance Summary Table for Luminance Error
Factors
df
Sum of Squares
Mean Squares
F
pvalue
Sample
56
0.266
0.005
22.356
Device
2
0.015
0.008
36.083
CTM
1
0.009
0.009
42.207
Sample*device
112
0.144
0.001
6.058
Sample*CTM
56
0.145
0.003
12.216
Device*CTM
2
0.001
0.001
2.372
0.098
Table 7.
Color Differences (), by Device, for 57 Samples, with Test Source
Device
40Dfe45
50Dfe45
5Dfe8
Mean
Mean
Mean
Calibration Method
± std
Maximum
± std
Maximum
± std
Maximum
sRGB
5.4
11.6
3.8
9.7
4.0
7.9
−1CF
(#16-yellow)
(#16-yellow)
(#17-magenta)
sRGB
4.0
10.6
3.6
11.2
3.1
8.1
−3CF
(#16-yellow)
(#16-yellow)
(5Y 8.5/12)
minXYZ
3.2
6.6
3.1
7.4
2.0
4.3
−1CF
(#7-orange)
(#15-red)
(#15-red)
minXYZ
1.9
5.9
1.7
5.7
1.7
4.3
−3CF
(5G 7/10)
(#15-red)
(#15-red)
Table 8.
Analysis of Variance Summary Table for Color Difference
Factors
df
Sum of Squares
Mean Squares
F
pvalue
Sample
56
1052.035
18.786
22.217
Device
2
101.377
50.689
59.944
Re-scaling
1
139.924
139.924
165.473
CTM
1
495.801
495.801
586.329
Sample*device
112
185.262
1.654
1.956
Sample*re-scaling
56
163.773
2.925
3.459
Sample*CTM
56
501.003
8.946
10.580
Device*re-scaling
2
16.869
8.434
9.974
Device*CTM
2
20.448
10.224
12.091
Re-scaling*CTM
1
1.599
1.599
1.891
0.170
Table 9.
Overview Table of Error Percentages in Luminance and Color Differences () Observed in the Literature and Experimental Conditions
Fisheye lens.
Not specified.
CTM determined using an ultra-cool daylight source (CCT not specified).
CTM determined using a cool incandescent source (5050K) + 3CF.
Ultra-cool daylight + real room lighting.
HDR assembly part of Debevec and Malik’s method [10].
CTM determined at 5500K.