Gabriele Simone, Marco Gaiani, Andrea Ballabeni, and Alessandro Rizzi, "Complex process of image color correction: a test of a target-based framework," J. Opt. Soc. Am. A 38, 663-674 (2021)
This paper aims at presenting the complexity of the process of image target-based color correction (CC). We present issues encountered from acquisition to rendering using colorimetric traditional tools. Target-based CC can be seen as an optimization problem. We have tested SAT and HUE adaptive fine tuning (SHAFT) an automated framework for target-based CC. A key element of SHAFT is an iterative CIEDE2000 variation comparison between a reference and target image. In this work we replace the standard CIEDE2000 with the Euclidean color-difference formula for small–medium color differences in log-compressed Optical Society of America’s Committee on Uniform Color Scales (OSA-UCS) space. Results are presented using both formulae. A discussion on the complexity of scene color departures and correction performances concludes the paper. The effect of real scene complexity is shown and how colors are subject to disordered shifts in the color space. Because of this complexity, the role of the CC method as a different color error minimizer emerges.
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.
T Test on processing time: 0.356617002T Test on number of cycles: 0.889405395
For each image and method ($\Delta {E_{00}}$-based and $\Delta {E_E}$-based), the measurements listed in Section 5 were performed, using the dataset of 22 XCC RAW images illustrated in Fig. 2.
Table 2.
Number of XCC Image Preferences for Each Observer
Observer
S1
S2
S3
S4
S5
S6
S7
S8
S9
S10
S11
S12
Total
11
10
13
8
10
6
10
12
10
7
11
12
120
8
9
6
11
9
13
9
7
9
12
8
7
108
Table 3.
Mean Opinion Score Corrected with the -Score Technique per Each Observer on the 72 Patches (36 for Each Color Space)a
Observer
S1
S2
S3
S4
S5
S6
S7
S8
S9
S10
S11
S12
Total
–0.27
0.5
0.58
0.23
0.42
0.11
0.22
0.56
0.03
–0.3
0.49
0.59
–0.21
0.26
0.12
0.03
0.83
Average $\Delta {E_{00}}$ and $\Delta {E_E}$ scores are represented as $Z$-scores for each experimental subject.
Table 4.
Mean Opinion Score Corrected with the -Score Technique of All Observers per Each Patch (36 for Each Color Space)a
Patch Name
1
7
13
14
3
15
16
11
17
Total
0.21
–0.77
0.16
0.33
0.77
0.63
0.03
–0.2
–0.8
0.17
0.15
0.76
0.62
Average $\Delta {E_{00}}$ and $\Delta {E_E}$ scores are represented as $Z$-scores for each experimental subject.
Factor is represented by the different development, based on $\Delta {E_{00}}$ and $\Delta {E_E}$. MS represents the mean of square, DoF represents the degrees of freedom, $F$ is Fisher value, ${F_{\rm{crit}}}$ represents the critical value of $F$ distribution. The two groups represent the $\Delta {E_{00}}$ and $\Delta {E_E}$ methods. The significance level obtained (0.480845) is below 0.05 (${\rm{alpha}} = {0.05}$), meaning that there is no significant difference between the means of the two groups.
Table 6.
Extended Experiment B: Mean Opinion Score Corrected with the -Score Technique per Each Observer on the 168 Patches Observed (84 for Each Color Space)a
Observer
S1
S2
S3
S4
S5
S6
S7
S8
S9
S10
S11
S12
Total
0.17
0.52
0.27
0.39
0.72
0.17
0.22
0.35
–0.6
–0.73
0.22
0.52
0.3
0.44
0.73
0.17
0.31
0.61
–0.61
–0.7
0.02
Average $\Delta {E_{00}}$ and $\Delta {E_E}$ scores are represented as $Z$-scores for each experimental subject.
Table 7.
Extended Experiment B: Mean Opinion Score Corrected with the -Score Technique of All Observers per Each Patch (84 for Each Color Space)a
Patch Name
13
3
17
5
6
10
8
Total
0.03
0.09
0.87
0.05
–0.01
0.09
0.95
0.1
0.02
Average $\Delta {E_{00}}$ and $\Delta {E_E}$ scores are represented as $Z$-scores for each patch.
Factor is represented by the different CC solutions, based on $\Delta {E_{00}}$ and $\Delta {E_E}$. MS represents the mean of square, DoF represents the degrees of freedom, $F$ is Fisher value, ${{{F}}_{\rm{crit}}}$ represents the critical value of $F$ distribution. The significance level obtained (0.014405) is below 0.05 (${\rm{alpha}} = {0.05}$).
Table 9.
Quantitative Analysis of the Influence of Increasing Glare and Noise on the Two Techniques Based on and Measured Using Units
Acquired, CIEDE2000 and ${{{\Delta}}E_{E}}$ from left to right, respectively. Reference values are ${\rm{L}}^* = {20.64}$, ${\rm{a}}^* = {0.13}$, ${\rm{b}}^* = - {0.46}$.
Acquired, CIEDE2000 and ${{{\Delta}}E_{E}}$ from left to right, respectively. In italic values ${\lt}{{1}}$ and in bold-italic values ${\gt}\!{{10}}$, respectively.
Table 12.
Lightness Ratio between the White Patch (#19) and the Black Patch (#24)a
Image
Acquired
_48_R01_
5.61
4.12
4.11
_DSC0134
3.90
4.45
4.44
_DSC0136
7.89
4.99
5.00
_DSC0945
4.25
3.48
3.48
_DSC2118
3.99
4.30
4.30
_DSC2416
4.40
4.41
4.40
_DSC2524
4.28
4.62
4.62
_DSC2801
3.65
4.69
4.68
_DSC2935
4.59
3.73
3.73
_DSC3197
5.41
4.58
4.56
_DSC3630
2.73
4.48
4.46
_DSC6062
6.00
4.11
4.11
_MG_9935
4.62
4.11
4.11
_T4A4626
6.70
4.24
4.25
_T4A4846
5.99
4.29
4.25
_T4A5900
4.68
4.30
4.70
_wb_L5_A
3.98
4.92
4.92
1__200Is
4.20
4.75
4.74
2__200Is
3.91
4.75
4.74
3__200Is
3.90
4.75
4.75
APC_0029
4.45
3.89
3.88
APC_0033
4.02
3.52
3.52
Acquired, CIEDE2000 and ${{{\Delta}}E_{E}}$ from left to right, respectively. In bold values ${\lt}{{4}}$ and in bold-italic values ${\gt}{{5}}$, respectively. The reference value is 4.61.
T Test on processing time: 0.356617002T Test on number of cycles: 0.889405395
For each image and method ($\Delta {E_{00}}$-based and $\Delta {E_E}$-based), the measurements listed in Section 5 were performed, using the dataset of 22 XCC RAW images illustrated in Fig. 2.
Table 2.
Number of XCC Image Preferences for Each Observer
Observer
S1
S2
S3
S4
S5
S6
S7
S8
S9
S10
S11
S12
Total
11
10
13
8
10
6
10
12
10
7
11
12
120
8
9
6
11
9
13
9
7
9
12
8
7
108
Table 3.
Mean Opinion Score Corrected with the -Score Technique per Each Observer on the 72 Patches (36 for Each Color Space)a
Observer
S1
S2
S3
S4
S5
S6
S7
S8
S9
S10
S11
S12
Total
–0.27
0.5
0.58
0.23
0.42
0.11
0.22
0.56
0.03
–0.3
0.49
0.59
–0.21
0.26
0.12
0.03
0.83
Average $\Delta {E_{00}}$ and $\Delta {E_E}$ scores are represented as $Z$-scores for each experimental subject.
Table 4.
Mean Opinion Score Corrected with the -Score Technique of All Observers per Each Patch (36 for Each Color Space)a
Patch Name
1
7
13
14
3
15
16
11
17
Total
0.21
–0.77
0.16
0.33
0.77
0.63
0.03
–0.2
–0.8
0.17
0.15
0.76
0.62
Average $\Delta {E_{00}}$ and $\Delta {E_E}$ scores are represented as $Z$-scores for each experimental subject.
Factor is represented by the different development, based on $\Delta {E_{00}}$ and $\Delta {E_E}$. MS represents the mean of square, DoF represents the degrees of freedom, $F$ is Fisher value, ${F_{\rm{crit}}}$ represents the critical value of $F$ distribution. The two groups represent the $\Delta {E_{00}}$ and $\Delta {E_E}$ methods. The significance level obtained (0.480845) is below 0.05 (${\rm{alpha}} = {0.05}$), meaning that there is no significant difference between the means of the two groups.
Table 6.
Extended Experiment B: Mean Opinion Score Corrected with the -Score Technique per Each Observer on the 168 Patches Observed (84 for Each Color Space)a
Observer
S1
S2
S3
S4
S5
S6
S7
S8
S9
S10
S11
S12
Total
0.17
0.52
0.27
0.39
0.72
0.17
0.22
0.35
–0.6
–0.73
0.22
0.52
0.3
0.44
0.73
0.17
0.31
0.61
–0.61
–0.7
0.02
Average $\Delta {E_{00}}$ and $\Delta {E_E}$ scores are represented as $Z$-scores for each experimental subject.
Table 7.
Extended Experiment B: Mean Opinion Score Corrected with the -Score Technique of All Observers per Each Patch (84 for Each Color Space)a
Patch Name
13
3
17
5
6
10
8
Total
0.03
0.09
0.87
0.05
–0.01
0.09
0.95
0.1
0.02
Average $\Delta {E_{00}}$ and $\Delta {E_E}$ scores are represented as $Z$-scores for each patch.
Factor is represented by the different CC solutions, based on $\Delta {E_{00}}$ and $\Delta {E_E}$. MS represents the mean of square, DoF represents the degrees of freedom, $F$ is Fisher value, ${{{F}}_{\rm{crit}}}$ represents the critical value of $F$ distribution. The significance level obtained (0.014405) is below 0.05 (${\rm{alpha}} = {0.05}$).
Table 9.
Quantitative Analysis of the Influence of Increasing Glare and Noise on the Two Techniques Based on and Measured Using Units
Acquired, CIEDE2000 and ${{{\Delta}}E_{E}}$ from left to right, respectively. Reference values are ${\rm{L}}^* = {20.64}$, ${\rm{a}}^* = {0.13}$, ${\rm{b}}^* = - {0.46}$.
Acquired, CIEDE2000 and ${{{\Delta}}E_{E}}$ from left to right, respectively. In italic values ${\lt}{{1}}$ and in bold-italic values ${\gt}\!{{10}}$, respectively.
Table 12.
Lightness Ratio between the White Patch (#19) and the Black Patch (#24)a
Image
Acquired
_48_R01_
5.61
4.12
4.11
_DSC0134
3.90
4.45
4.44
_DSC0136
7.89
4.99
5.00
_DSC0945
4.25
3.48
3.48
_DSC2118
3.99
4.30
4.30
_DSC2416
4.40
4.41
4.40
_DSC2524
4.28
4.62
4.62
_DSC2801
3.65
4.69
4.68
_DSC2935
4.59
3.73
3.73
_DSC3197
5.41
4.58
4.56
_DSC3630
2.73
4.48
4.46
_DSC6062
6.00
4.11
4.11
_MG_9935
4.62
4.11
4.11
_T4A4626
6.70
4.24
4.25
_T4A4846
5.99
4.29
4.25
_T4A5900
4.68
4.30
4.70
_wb_L5_A
3.98
4.92
4.92
1__200Is
4.20
4.75
4.74
2__200Is
3.91
4.75
4.74
3__200Is
3.90
4.75
4.75
APC_0029
4.45
3.89
3.88
APC_0033
4.02
3.52
3.52
Acquired, CIEDE2000 and ${{{\Delta}}E_{E}}$ from left to right, respectively. In bold values ${\lt}{{4}}$ and in bold-italic values ${\gt}{{5}}$, respectively. The reference value is 4.61.