Multispectral Data Correction Method for Painted Cultural Relics Under Non-Uniform Illumination Based on Prior Feature Constraints

Jiachen Li; Ke Wang; Huiqin Wang; Zhan Wang; Peize Han

doi:10.3788/LOP241890

Laser & Optoelectronics Progress, Volume. 62, Issue 8, 0830001(2025)

Multispectral Data Correction Method for Painted Cultural Relics Under Non-Uniform Illumination Based on Prior Feature Constraints

Jiachen Li^1、*, Ke Wang¹, Huiqin Wang¹, Zhan Wang², and Peize Han³

Author Affiliations

¹College of Information and Control Engineering, Xi'an University of Architecture and Technology, Xi'an 710055, Shaanxi , China

²Shaanxi Institute for the Preservation of Cultural Heritage, Xi'an 710075, Shaanxi , China

³College of Life Science and Technology, Beijing University of Chemical Technology, Beijing 100029, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(26)

Fig. 1. Framework diagram for non-uniformity illumination correction of multispectral imaging data

Download full size

Fig. 2. Schematic diagram of illumination component area splitting

Download full size

Fig. 3. Adaptive gamma corrected results and their grey scale histograms. (a) Adaptive gamma corrected image; (b) grey scale histogram

Download full size

Fig. 4. Schematic diagram of multispectral imaging system

Download full size

Fig. 5. Imaging results under non-uniform illumination. (a) 16-channel multispectral split-shot imaging; (b) 680-channel split-shot imaging of A1 and A2 regions; (c) split-shot grey scale histograms

Download full size

Fig. 6. Correction effects of different methods on simulated murals (region A1). (a) Pre-calibration image; (b) AGCWD; (c) LIME; (d) DUAL; (e) ATF; (f) RetinexNet; (g) CPGA-Net; (h) proposed method

Download full size

Fig. 7. Correction effects of different methods on simulated murals (region B2). (a) Pre-calibration image; (b) AGCWD; (c) LIME; (d) DUAL; (e) ATF; (f) RetinexNet; (g) CPGA-Net; (h) proposed method

Download full size

Fig. 8. Schematic diagram of iterative process of pigment sampling region and loss function minimization

Download full size

Fig. 9. Comparison of reflectance spectra before and after pigment correction. (a) Mineral green; (b) vermilion; (c) lazurite; (d) azurite; (e) ochre; (f) cinnabar

Download full size

Fig. 10. Analogue mural truth map. (a) Truth map for region A1; (b) truth map for region A2

Download full size

Fig. 11. Schematic diagram of experimental data of mural paintings of Dule Temple in separate shots

Download full size

Fig. 12. D5 and D6 split shot mural correction effect. (a) Before calibration; (b) after calibration

Download full size

Fig. 13. Schematic diagram of pigment sampling area. (a) Sampling map points in D5 region; (b) sampling points in D6 region

Download full size

Fig. 14. SVM classification results before and after mural correction. (a) Before calibration; (b) after calibration

Download full size

Fig. 15. Schematic diagram of experimental data and marker points for the Thirteenth Venerable

Download full size

Fig. 16. SVM classification results before and after the Thirteenth Venerable correction. (a) Before calibration; (b) after calibration

Download full size

Table 1. Sample multispectral image of simulated murals
View table
Table 1. Sample multispectral image of simulated murals
Number Category Number of samples
1 Mineral green 9336
2 Vermilion 10566
3 Lazurite 9963
4 Azurite 5060
5 Ochre 426
6 Cinnabar 4701

Table 2. RMSE metrics for different algorithmic calibration results

View table

Table 2. RMSE metrics for different algorithmic calibration results

Algorithm	Mineral green	Vermilion	Lazurite	Azurite	Ochre	Cinnabar
Before calibration	28.577	55.558	41.490	32.919	77.291	94.955
AGCWD	41.697	87.740	67.079	53.971	121.830	139.122
LIME	80.951	104.353	98.057	96.846	127.140	135.207
DUAL	38.159	49.165	39.406	42.294	56.741	66.874
ATF	121.983	136.750	137.359	142.362	157.268	159.905
RetinexNet	145.941	135.973	140.004	157.600	146.878	149.021
CPGA-Net	131.960	137.209	139.070	146.941	155.338	158.283
Proposed	13.959	22.591	23.269	30.210	41.054	37.916

Table 3. SAM metrics for different algorithmic calibration results

View table

Table 3. SAM metrics for different algorithmic calibration results

Algorithm	Mineral green	Vermilion	Lazurite	Azurite	Ochre	Cinnabar
Before calibration	0.814	0.736	0.608	0.606	0.595	0.750
AGCWD	0.868	0.785	0.538	0.562	0.606	0.796
LIME	0.725	0.648	0.419	0.518	0.488	0.634
DUAL	0.742	0.658	0.403	0.486	0.508	0.659
ATF	0.620	0.642	0.313	0.404	0.457	0.618
RetinexNet	0.610	0.499	0.284	0.500	0.376	0.516
CPGA-Net	0.630	0.544	0.313	0.469	0.407	0.556
Proposed	0.219	0.243	0.137	0.140	0.314	0.313

Table 4. SCM metrics for different algorithmic calibration results

View table

Table 4. SCM metrics for different algorithmic calibration results

Algorithm	Mineral Green	Vermilion	Lazurite	Azurite	Ochre	Cinnabar
Before calibration	0.685	0.740	0.820	0.821	0.827	0.731
AGCWD	0.645	0.706	0.858	0.845	0.821	0.698
LIME	0.748	0.796	0.913	0.868	0.883	0.805
DUAL	0.736	0.791	0.919	0.883	0.873	0.790
ATF	0.813	0.800	0.951	0.919	0.897	0.815
RetinexNet	0.819	0.877	0.959	0.877	0.929	0.869
CPGA-Net	0.807	0.855	0.951	0.891	0.918	0.849
Proposed	0.998	0.991	0.993	0.996	0.961	0.983

Table 5. Classification results of different methods before and after light correction (region A1)
View table
Table 5. Classification results of different methods before and after light correction (region A1)
Algorithm SVM SAM MLP MD ML SID
Beforecalibration
Aftercalibration

Table 6. Classification results of different methods before and after light correction (region B2)
View table
Table 6. Classification results of different methods before and after light correction (region B2)
Algorithm SVM SAM MLP MD ML SID
Beforecalibration
Aftercalibration

Table 7. Classification accuracy of different methods before and after illumination correction (region A1)

View table

Table 7. Classification accuracy of different methods before and after illumination correction (region A1)

Class	SVM				MLP		MD		ML		SID
Class	Before calibration	After calibration	Before calibration	After calibration	Before calibration	After calibration	Before calibration	After calibration	Before calibration	After calibration	Before calibration	After calibration
Mineral green	97.90%	97.14%	95.17%	96.82%	98.71%	98.33%	97.47%	96.26%	99.76%	99.70%	76.96%	95.82%
Vermilion	78.98%	97.89%	99.64%	99.89%	99.07%	98.13%	98.05%	98.60%	98.49%	99.37%	89.32%	99.85%
Lazurite	98.50%	99.57%	99.86%	99.52%	99.53%	99.62%	99.00%	99.45%	99.81%	99.90%	99.72%	99.25%
OA /%	90.31	98.04	97.96	98.62	94.06	98.66	89.53	97.37	96.90	98.83	90.19	98.19
Kappa× 100	84.08	96.57	96.37	97.55	89.60	97.63	82.81	95.38	94.53	97.93	83.53	96.79

Table 8. Classification accuracy of different methods before and after illumination correction (region B2)

View table

Table 8. Classification accuracy of different methods before and after illumination correction (region B2)

Class	SVM		SAM		MLP		MD		ML		SID
Class	Before calibration	After calibration	Before calibration	After calibration	Before calibration	After calibration	Before calibration	After calibration	Before calibration	After calibration	Before calibration	After calibration
Mineral green	99.43%	98.32%	95.53%	99.03%	99.67%	99.80%	98.85%	96.91%	99.78%	99.89%	85.74%	98.63%
Vermilion	83.10%	89.71%	94.87%	93.59%	96.80%	98.37%	96.48%	97.01%	99.07%	99.26%	92.34%	92.33%
Lazurite	98.85%	98.58%	99.48%	99.59%	96.02%	98.88%	89.08%	93.58%	99.68%	99.60%	96.66%	98.45%
Azurite	99.67%	99.51%	99.72%	99.95%	99.91%	99.86%	99.85%	99.46%	99.04%	99.73%	99.48%	99.57%
Ochre	88.71%	92.35%	99.45%	99.95%	97.45%	98.28%	95.95%	97.92%	96.11%	96.51%	99.23%	99.50%
Cinnabar	60.29%	79.27%	96.04%	97.82%	96.08%	97.49%	69.8%	96.97%	98.97%	98.92%	96.14%	97.98%
OA /%	90.83	95.57	97.09	98.58	96.87	98.37	92.74	95.79	98.58	97.96	96.20	98.00
Kappa× 100	88.19	92.73	96.41	98.08	95.21	97.80	90.33	94.33	98.07	96.88	94.86	97.29

Table 9. Results of ablation experiments
View table
Table 9. Results of ablation experiments
Model RMSE SAM SCM
No area adaptive illumination correction module 42.363 0.301 0.925
No global calibration module for a priori support 34.936 0.403 0.787
Proposed 28.166 0.175 0.987

Table 10. Indicators of results of calibration of proposed method

View table

Table 10. Indicators of results of calibration of proposed method

Sampling Region	Before calibration			After calibration
Sampling Region	RMSE	SAM	SCM		RMSE	SAM	SCM
D5-1	55.345	0.158	0.563	36.131		0.134	0.741
D5-2	59.625	0.168	0.704	45.058		0.146	0.817
D5-3	60.905	0.178	0.706	41.179		0.156	0.816
D5-4	52.372	0.236	0.733	34.169		0.159	0.884
D6-1	56.673	0.322	0.399	41.315		0.204	0.474
D6-2	58.642	0.429	0.477	46.903		0.205	0.677
D6-3	71.820	0.388	0.445	62.567		0.233	0.504

Tools

Get Citation

Copy Citation Text

Jiachen Li, Ke Wang, Huiqin Wang, Zhan Wang, Peize Han. Multispectral Data Correction Method for Painted Cultural Relics Under Non-Uniform Illumination Based on Prior Feature Constraints[J]. Laser & Optoelectronics Progress, 2025, 62(8): 0830001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Spectroscopy

Received: Aug. 22, 2024

Accepted: Oct. 28, 2024

Published Online: Apr. 7, 2025

The Author Email:

DOI:10.3788/LOP241890

CSTR:32186.14.LOP241890

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Sample multispectral image of simulated murals

Table 1. Sample multispectral image of simulated murals

Table 2. RMSE metrics for different algorithmic calibration results

Table 2. RMSE metrics for different algorithmic calibration results

Table 3. SAM metrics for different algorithmic calibration results

Table 3. SAM metrics for different algorithmic calibration results

Table 4. SCM metrics for different algorithmic calibration results

Table 4. SCM metrics for different algorithmic calibration results

Table 5. Classification results of different methods before and after light correction (region A1)

Table 5. Classification results of different methods before and after light correction (region A1)

Table 6. Classification results of different methods before and after light correction (region B2)

Table 6. Classification results of different methods before and after light correction (region B2)

Table 7. Classification accuracy of different methods before and after illumination correction (region A1)

Table 7. Classification accuracy of different methods before and after illumination correction (region A1)

Table 8. Classification accuracy of different methods before and after illumination correction (region B2)

Table 8. Classification accuracy of different methods before and after illumination correction (region B2)

Table 9. Results of ablation experiments

Table 9. Results of ablation experiments

Table 10. Indicators of results of calibration of proposed method

Table 10. Indicators of results of calibration of proposed method