Ville Heikkinen, Clara Cámara, Tapani Hirvonen, and Niko Penttinen, "Spectral imaging using consumer-level devices and kernel-based regression," J. Opt. Soc. Am. A 33, 1095-1110 (2016)
Hyperspectral reflectance factor image estimations were performed in the 400–700 nm wavelength range using a portable consumer-level laptop display as an adjustable light source for a trichromatic camera. Targets of interest were ColorChecker Classic samples, Munsell Matte samples, geometrically challenging tempera icon paintings from the turn of the 20th century, and human hands. Measurements and simulations were performed using Nikon D80 RGB camera and Dell Vostro 2520 laptop screen as a light source. Estimations were performed without spectral characteristics of the devices and by emphasizing simplicity for training sets and estimation model optimization. Spectral and color error images are shown for the estimations using line-scanned hyperspectral images as the ground truth. Estimations were performed using kernel-based regression models via a first-degree inhomogeneous polynomial kernel and a Matérn kernel, where in the latter case the median heuristic approach for model optimization and link function for bounded estimation were evaluated. Results suggest modest requirements for a training set and show that all estimation models have markedly improved accuracy with respect to the DE00 color distance (up to 99% for paintings and hands) and the Pearson distance (up to 98% for paintings and 99% for hands) from a weak training set (Digital ColorChecker SG) case when small representative training data were used in the estimation.
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.
Estimation Results for the ColorChecker Classic When Using the DSG as the Training Set and Nikon D80 RGB Values and One, Two, and Four Light Sourcesa
Method
RMSE
PD
(D65)
(A)
(F11)
Three Bands (Lights: W)
PBM
0.04302/0.0746/0.0808
0.01582/0.0418/0.1113
1.93/3.24/3.46
3.06/6.35/9.45
2.36/4.01/6.33
MBM (CVGS)
0.02534/0.0509/0.0518
0.00493/0.0196/0.0222
1.69/3.08/3.08
2.09/3.50/5.59
1.80/2.77/2.92
MBM (MH)
0.02475/0.0479/0.0515
0.00496/0.0201/0.0237
1.69/3.33/3.41
2.06/3.52/6.06
1.76/2.69/3.16
MLBM (CVGS)
0.02388/0.0449/0.0506
0.00421/0.0119/0.0177
1.59/3.10/4.05
1.89/4.14/5.26
1.72/3.01/3.71
MLBM (MH)
0.02378/0.0475/0.0508
0.00437/0.0123/0.0190
1.65/3.30/4.44
1.93/4.55/6.02
1.76/3.08/4.10
Six Bands (Lights: W,G)
PBM
0.02689/0.0432/0.0436
0.00516/0.0219/0.0267
1.55/2.91/3.25
1.63/2.54/3.33
1.85/3.43/3.81
MBM (CVGS)
0.02128/0.0451/0.0455
0.00301/0.0109/0.0163
1.64/3.08/3.48
1.68/2.66/3.21
1.76/2.49/3.18
MBM (MH)
0.02065/0.0359/0.0475
0.00312/0.0115/0.0181
1.68/3.19/3.23
1.84/3.31/3.84
1.78/2.60/3.32
MLBM (CVGS)
0.02169/0.0357/0.0484
0.00341/0.0131/0.0183
1.79/3.21/3.85
1.82/3.96/4.01
1.91/2.85/3.94
MLBM (MH)
0.02089/0.0328/0.0477
0.00330/0.0147/0.0148
1.81/2.89/3.97
1.94/4.05/4.06
1.97/3.12/4.11
11 Bands (Lights: W,R,G,B)
PBM
0.02811/0.0464/0.0495
0.00539/0.0226/0.0292
1.61/3.29/3.48
1.78/3.10/3.84
1.74/3.22/3.30
MBM (CVGS)
0.02356/0.0417/0.0435
0.00351/0.0110/0.0246
1.68/3.20/3.56
1.88/3.48/3.71
1.79/2.76/3.21
MBM (MH)
0.02097/0.0355/0.0388
0.00338/0.0140/0.0259
1.60/3.04/3.30
1.81/3.46/3.63
1.72/2.49/3.40
MLBM (CVGS)
0.02250/0.0344/0.0380
0.00309/0.0131/0.0135
1.64/2.79/3.65
1.77/3.50/3.77
1.79/2.85/3.71
MLBM (MH)
0.02083/0.0329/0.0379
0.00314/0.0125/0.0179
1.68/2.48/4.06
1.82/3.07/4.18
1.87/3.07/4.24
Spectral and color errors for the PBM, MBM, and MLBM. The median heuristic optimization is denoted with MH and the 10-fold CV grid-based optimization is denoted with CVGS. For each error metric the results correspond to Average/95th percentile/Maximum values. For each lighting case (W), (W,G), (W,R,G,B) the best result is boldfaced.
Table 2.
Estimation Results for the Munsell Matte Collection Using the DSG as the Training Set and Nikon D80 RGB Values and One, Two, and Four Light Sourcesa
Method
RMSE
PD
(D65)
(A)
(F11)
Three Bands (Lights: W)
PBM
0.06930/0.1194/0.1712
0.01416/0.0505/0.1491
4.38/6.30/8.26
5.04/7.41/10.09
4.67/7.27/11.39
MBM (CVGS)
0.06341/0.1187/0.1966
0.00670/0.0261/0.1232
4.44/6.36/8.32
4.76/6.81/9.13
4.60/6.77/10.14
MBM (MH)
0.06044/0.1067/0.1548
0.00678/0.0265/0.1207
4.40/6.34/8.46
4.69/6.84/9.24
4.56/6.68/10.45
MLBM (CVGS)
0.06171/0.1143/0.1858
0.00591/0.0218/0.1689
4.32/6.33/8.35
4.62/6.83/9.14
4.49/6.73/10.68
MLBM (MH)
0.06024/0.1080/0.1597
0.00604/0.0226/0.1662
4.36/6.40/8.51
4.63/6.93/9.28
4.52/6.77/10.86
Six Bands (Lights: W,G)
PBM
0.05718/0.1067/0.1509
0.00500/0.0132/0.1673
4.30/6.05/8.02
4.45/6.27/8.45
4.63/6.52/9.21
MBM (CVGS)
0.05800/0.1201/0.2192
0.00415/0.0147/0.1186
4.38/6.39/8.22
4.58/6.45/8.64
4.55/6.55/9.40
MBM (MH)
0.05604/0.1018/0.1380
0.00480/0.0169/0.1014
4.37/6.31/8.29
4.60/6.47/8.61
4.56/6.52/9.49
MLBM (CVGS)
0.05737/0.1147/0.2061
0.00373/0.0109/0.1249
4.52/6.55/8.51
4.67/6.71/8.41
4.65/6.67/9.11
MLBM (MH)
0.05641/0.1044/0.1508
0.00407/0.0121/0.1158
4.52/6.46/8.26
4.73/6.79/8.45
4.70/6.70/9.20
11 Bands (Lights: W,R,G,B)
PBM
0.05311/0.0981/0.1430
0.00381/0.0093/0.1745
4.03/5.87/7.67
4.16/6.03/8.12
4.21/6.13/8.88
MBM (CVGS)
0.05855/0.1172/0.2233
0.00350/0.0101/0.1269
4.39/6.31/8.28
4.58/6.41/8.72
4.39/6.25/8.74
MBM (MH)
0.05802/0.1042/0.1462
0.00430/0.0142/0.1053
4.52/6.46/8.66
4.74/6.65/8.89
4.74/6.74/9.50
MLBM (CVGS)
0.05577/0.1104/0.1983
0.00304/0.0083/0.1441
4.31/6.38/8.42
4.47/6.46/8.65
4.40/6.33/9.11
MLBM (MH)
0.05813/0.1064/0.1541
0.00362/0.0094/0.1221
4.55/6.47/8.65
4.77/6.74/8.71
4.76/6.84/9.22
Spectral and color errors for the PBM, MBM, and MLBM. The median heuristic optimization is denoted with MH and the 10-fold CV grid-based optimization is denoted with CVGS. For each error metric the results correspond to Average/95th percentile/Maximum values. For each lighting case (W), (W,G), (W,R,G,B) the best result is boldfaced.
Tables (2)
Table 1.
Estimation Results for the ColorChecker Classic When Using the DSG as the Training Set and Nikon D80 RGB Values and One, Two, and Four Light Sourcesa
Method
RMSE
PD
(D65)
(A)
(F11)
Three Bands (Lights: W)
PBM
0.04302/0.0746/0.0808
0.01582/0.0418/0.1113
1.93/3.24/3.46
3.06/6.35/9.45
2.36/4.01/6.33
MBM (CVGS)
0.02534/0.0509/0.0518
0.00493/0.0196/0.0222
1.69/3.08/3.08
2.09/3.50/5.59
1.80/2.77/2.92
MBM (MH)
0.02475/0.0479/0.0515
0.00496/0.0201/0.0237
1.69/3.33/3.41
2.06/3.52/6.06
1.76/2.69/3.16
MLBM (CVGS)
0.02388/0.0449/0.0506
0.00421/0.0119/0.0177
1.59/3.10/4.05
1.89/4.14/5.26
1.72/3.01/3.71
MLBM (MH)
0.02378/0.0475/0.0508
0.00437/0.0123/0.0190
1.65/3.30/4.44
1.93/4.55/6.02
1.76/3.08/4.10
Six Bands (Lights: W,G)
PBM
0.02689/0.0432/0.0436
0.00516/0.0219/0.0267
1.55/2.91/3.25
1.63/2.54/3.33
1.85/3.43/3.81
MBM (CVGS)
0.02128/0.0451/0.0455
0.00301/0.0109/0.0163
1.64/3.08/3.48
1.68/2.66/3.21
1.76/2.49/3.18
MBM (MH)
0.02065/0.0359/0.0475
0.00312/0.0115/0.0181
1.68/3.19/3.23
1.84/3.31/3.84
1.78/2.60/3.32
MLBM (CVGS)
0.02169/0.0357/0.0484
0.00341/0.0131/0.0183
1.79/3.21/3.85
1.82/3.96/4.01
1.91/2.85/3.94
MLBM (MH)
0.02089/0.0328/0.0477
0.00330/0.0147/0.0148
1.81/2.89/3.97
1.94/4.05/4.06
1.97/3.12/4.11
11 Bands (Lights: W,R,G,B)
PBM
0.02811/0.0464/0.0495
0.00539/0.0226/0.0292
1.61/3.29/3.48
1.78/3.10/3.84
1.74/3.22/3.30
MBM (CVGS)
0.02356/0.0417/0.0435
0.00351/0.0110/0.0246
1.68/3.20/3.56
1.88/3.48/3.71
1.79/2.76/3.21
MBM (MH)
0.02097/0.0355/0.0388
0.00338/0.0140/0.0259
1.60/3.04/3.30
1.81/3.46/3.63
1.72/2.49/3.40
MLBM (CVGS)
0.02250/0.0344/0.0380
0.00309/0.0131/0.0135
1.64/2.79/3.65
1.77/3.50/3.77
1.79/2.85/3.71
MLBM (MH)
0.02083/0.0329/0.0379
0.00314/0.0125/0.0179
1.68/2.48/4.06
1.82/3.07/4.18
1.87/3.07/4.24
Spectral and color errors for the PBM, MBM, and MLBM. The median heuristic optimization is denoted with MH and the 10-fold CV grid-based optimization is denoted with CVGS. For each error metric the results correspond to Average/95th percentile/Maximum values. For each lighting case (W), (W,G), (W,R,G,B) the best result is boldfaced.
Table 2.
Estimation Results for the Munsell Matte Collection Using the DSG as the Training Set and Nikon D80 RGB Values and One, Two, and Four Light Sourcesa
Method
RMSE
PD
(D65)
(A)
(F11)
Three Bands (Lights: W)
PBM
0.06930/0.1194/0.1712
0.01416/0.0505/0.1491
4.38/6.30/8.26
5.04/7.41/10.09
4.67/7.27/11.39
MBM (CVGS)
0.06341/0.1187/0.1966
0.00670/0.0261/0.1232
4.44/6.36/8.32
4.76/6.81/9.13
4.60/6.77/10.14
MBM (MH)
0.06044/0.1067/0.1548
0.00678/0.0265/0.1207
4.40/6.34/8.46
4.69/6.84/9.24
4.56/6.68/10.45
MLBM (CVGS)
0.06171/0.1143/0.1858
0.00591/0.0218/0.1689
4.32/6.33/8.35
4.62/6.83/9.14
4.49/6.73/10.68
MLBM (MH)
0.06024/0.1080/0.1597
0.00604/0.0226/0.1662
4.36/6.40/8.51
4.63/6.93/9.28
4.52/6.77/10.86
Six Bands (Lights: W,G)
PBM
0.05718/0.1067/0.1509
0.00500/0.0132/0.1673
4.30/6.05/8.02
4.45/6.27/8.45
4.63/6.52/9.21
MBM (CVGS)
0.05800/0.1201/0.2192
0.00415/0.0147/0.1186
4.38/6.39/8.22
4.58/6.45/8.64
4.55/6.55/9.40
MBM (MH)
0.05604/0.1018/0.1380
0.00480/0.0169/0.1014
4.37/6.31/8.29
4.60/6.47/8.61
4.56/6.52/9.49
MLBM (CVGS)
0.05737/0.1147/0.2061
0.00373/0.0109/0.1249
4.52/6.55/8.51
4.67/6.71/8.41
4.65/6.67/9.11
MLBM (MH)
0.05641/0.1044/0.1508
0.00407/0.0121/0.1158
4.52/6.46/8.26
4.73/6.79/8.45
4.70/6.70/9.20
11 Bands (Lights: W,R,G,B)
PBM
0.05311/0.0981/0.1430
0.00381/0.0093/0.1745
4.03/5.87/7.67
4.16/6.03/8.12
4.21/6.13/8.88
MBM (CVGS)
0.05855/0.1172/0.2233
0.00350/0.0101/0.1269
4.39/6.31/8.28
4.58/6.41/8.72
4.39/6.25/8.74
MBM (MH)
0.05802/0.1042/0.1462
0.00430/0.0142/0.1053
4.52/6.46/8.66
4.74/6.65/8.89
4.74/6.74/9.50
MLBM (CVGS)
0.05577/0.1104/0.1983
0.00304/0.0083/0.1441
4.31/6.38/8.42
4.47/6.46/8.65
4.40/6.33/9.11
MLBM (MH)
0.05813/0.1064/0.1541
0.00362/0.0094/0.1221
4.55/6.47/8.65
4.77/6.74/8.71
4.76/6.84/9.22
Spectral and color errors for the PBM, MBM, and MLBM. The median heuristic optimization is denoted with MH and the 10-fold CV grid-based optimization is denoted with CVGS. For each error metric the results correspond to Average/95th percentile/Maximum values. For each lighting case (W), (W,G), (W,R,G,B) the best result is boldfaced.