Multi-Spectral Color Data Dimension Reduction Model Research Based on Sparse Representation

Xinyi Fang; Xiaoxia Wan; Shuo Shi; Xiao Teng; Junyan Yu

doi:10.3788/LOP202158.2230003

Journals >Laser & Optoelectronics Progress >Volume 58 >Issue 22 >Page 2230003 > Article

Laser & Optoelectronics Progress
Vol. 58, Issue 22, 2230003 (2021)

Multi-Spectral Color Data Dimension Reduction Model Research Based on Sparse Representation

Xinyi Fang¹, Xiaoxia Wan^1、*, Shuo Shi², Xiao Teng¹, and Junyan Yu¹

Author Affiliations

¹Color Science Laboratory, School of Printing and Packaging, Wuhan University, Wuhan, Hubei 430079, China

²State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan, Hubei 430079, China

show less

DOI: 10.3788/LOP202158.2230003 Cite this Article Set citation alerts

Xinyi Fang, Xiaoxia Wan, Shuo Shi, Xiao Teng, Junyan Yu. Multi-Spectral Color Data Dimension Reduction Model Research Based on Sparse Representation[J]. Laser & Optoelectronics Progress, 2021, 58(22): 2230003 Copy Citation Text

show less

RMSE distribution of each sample with SG color card as training set and Munsell color block as testset when each sample is reduced to six dimensions and reconstructed. (a) Spectral error distribution diagram obtained by PCA method; (b) spectral error distribution diagram obtained by SR method; (c) RMSE curve obtained by PCA method; (d) RMSE curve obtained by SR method

Fig. 1. R_MSE distribution of each sample with SG color card as training set and Munsell color block as testset when each sample is reduced to six dimensions and reconstructed. (a) Spectral error distribution diagram obtained by PCA method; (b) spectral error distribution diagram obtained by SR method; (c) R_MSE curve obtained by PCA method; (d) R_MSE curve obtained by SR method

Download full size

Partial sample spectral reflectance curve, and reconstructed spectral curve based on PCA method and SR method. (a) Samples with high fitting degree; (b) samples of general fitness; (c) samples with poor fit

Fig. 2. Partial sample spectral reflectance curve, and reconstructed spectral curve based on PCA method and SR method. (a) Samples with high fitting degree; (b) samples of general fitness; (c) samples with poor fit

Download full size

Fig. 3. Multispectral images from the University of Eastern Finland spectral image database. (a) Original image; (b) image reduced in dimension and reconstructed by SR method

Download full size

Dataset	Dimension			Method	R_MSE			G_FC /%
Dataset	Dimension			Method	Mean	Max	Min	Mean	Max	Min
Munsell-SG		6	PCA		0.017	0.048	0.003	99.825	99.999	98.066
		6	SR		0.011	0.039	0.002	99.884	99.998	97.905
		5	SR		0.011	0.039	0.002	99.884	99.998	97.905
			PCA		0.016	0.044	0.004	99.797	99.997	97.626
			SR		0.015	0.040	0.002	99.828	99.994	97.620
		4	SR		0.015	0.040	0.002	99.828	99.994	97.620
			PCA		0.020	0.068	0.004	99.626	99.995	93.961
			SR		0.019	0.068	0.003	99.648	99.994	93.779
		3	SR		0.019	0.068	0.003	99.648	99.994	93.779
			PCA		0.026	0.085	0.006	99.314	99.992	93.793
			SR		0.025	0.086	0.003	99.375	99.984	93.485
		SG-Munsell			0.025	0.086	0.003	99.375	99.984	93.485	6	PCA	0.068	0.655	0.005	99.298	99.993	76.455
SR				0.010	0.064	0.001	99.900	99.997	92.273		6
SR				0.010	0.064	0.001	99.900	99.997	92.273	5
PCA				0.051	0.323	0.005	99.450	99.993	69.969
SR				0.015	0.066	0.002	99.805	99.997	91.971
SR				0.015	0.066	0.002	99.805	99.997	91.971	4
PCA				0.047	0.253	0.005	99.383	99.993	77.736
SR				0.018	0.088	0.002	99.667	99.997	86.454
SR				0.018	0.088	0.002	99.667	99.997	86.454	3
PCA				0.041	0.102	0.004	99.335	99.986	85.781
SR				0.025	0.098	0.002	99.415	99.994	84.614

Table 1. Comparison of spectral reconstruction accuracy between SR method and PCA method

Dataset	Method	ΔE₀₀
Dataset	Method	Mean	Max	Min
Munsell-SG	PCA	1.517	9.904	0.047
Munsell-SG	SR	0.451	2.328	0.049
SG-Munsell	PCA	5.939	52.727	0.222
SG-Munsell	SR	0.430	4.675	0.011

Table 2. Comparison of chromaticity reconstruction accuracy between SR method and PCA method, under D65 light source

Dataset	Method	A	C	D65	D50
Munsell-SG	PCA	1.399	1.528	1.517	1.453
Munsell-SG	SR	0.277	0.501	0.451	0.373
SG-Munsell	PCA	6.251	5.972	5.939	6.072
SG-Munsell	SR	0.276	0.485	0.430	0.367

Table 3. Average color difference comparison under different light sources

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information