HSV Space-Based Nonlinear Adaptive Low-Light Image Enhancement Algorithm

Fig. 1. Processing effects of the algorithm of reference [11]. (a) Original images; (b) original image inversion; (c) results of the algorithm of reference [11]

Download full size

Fig. 2. Flow chart of the proposed algorithm

Download full size

Fig. 3. Comparison of adaptive gamma correction processing effects. (a) Original images; (b) conventional gamma correction; (c) adaptive gamma correction

Download full size

Fig. 4. Schematic illustration of the processing of the luminance adaptive contrast enhancement algorithm

Download full size

Fig. 5. Adaptive image fusion effects. (a) Original images; (b) proposed algorithm without the adaptive image fusion algorithm; (c) proposed algorithm

Download full size

Fig. 6. Comparison results of the LOL dataset. (a) Original image; (b) Dong algorithm; (c) KinD++ algorithm; (d) RetinexNet algorithm; (e) Retinexformer algorithm; (f) EnlightGAN algorithm; (g) Zero-DCE algorithm; (h) proposed algorithm

Download full size

Fig. 7. Comparison results of the DICM dataset. (a) Original image; (b) Dong algorithm; (c) KinD++ algorithm; (d) RetinexNet algorithm; (e) Retinexformer algorithm; (f) EnlightGAN algorithm; (g) Zero-DCE algorithm; (h) proposed algorithm

Download full size

Fig. 8. Comparison results of the LIME dataset. (a) Original image; (b) Dong algorithm; (c) KinD++ algorithm; (d) RetinexNet algorithm; (e) Retinexformer algorithm; (f) EnlightGAN algorithm; (g) Zero-DCE algorithm; (h) proposed algorithm

Download full size

Fig. 9. Comparison results of the NPE dataset. (a) Original image; (b) Dong algorithm; (c) KinD++ algorithm; (d) RetinexNet algorithm; (e) Retinexformer algorithm; (f) EnlightGAN algorithm; (g) Zero-DCE algorithm; (h) proposed algorithm

Download full size

Fig. 10. Comparison of the enhancement effects of different modules on low-light images. (a) Original images; (b) ABA; (c) -w/o BACE; (d) -w/o AIF; (e) proposed algorithm

Download full size

Fig. 11. Enhancement effects of different gamma values on images with different gray mean values. (a) Original images; (b) $γ = 2$ ; (c) $γ = 6$ ; (d) $γ = 12$ ; (e) $γ = 20$

Download full size

Fig. 12. Enhancement effects of different gamma values on images with different gray mean values. (a) $R_{G M} \in [0, 25)$ ; (b) $R_{G M} \in [25, 100)$ ;(c) $R_{G M} \in [100, 256)$

Download full size

Table 1. PSNR evaluation results of different algorithms

View table

Table 1. PSNR evaluation results of different algorithms

Algorithm	LOL	DICM	LIME	NPE	Average
Dong	10.4246	13.5263	11.8872	12.5767	12.1037
KinD++	7.8411	13.3278	11.9120	12.8265	11.4769
RetinexNet	8.4378	11.6582	9.9230	11.3168	10.3340
Retinexformer	7.7975	10.4485	13.1782	12.5196	10.9860
EnlightGAN	10.2751	12.7715	10.6697	13.4907	11.8017
Zero-DCE	13.0488	14.6403	12.7626	13.8699	13.5804
Proposed algorithm	15.8907	17.2655	15.2657	16.5140	16.2340

Table 2. SSIM evaluation results of different algorithms

View table

Table 2. SSIM evaluation results of different algorithms

Algorithm	LOL	DICM	LIME	NPE	Average
Dong	0.1597	0.4891	0.3878	0.5733	0.4025
KinD++	0.1565	0.5001	0.3222	0.5469	0.3816
RetinexNet	0.1169	0.4363	0.4559	0.5860	0.3988
Retinexformer	0.1770	0.3924	0.4577	0.5547	0.3955
EnlightGAN	0.1795	0.4845	0.3618	0.6494	0.4188
Zero-DCE	0.2640	0.5280	0.4232	0.6340	0.4623
Proposed algorithm	0.3114	0.5762	0.4940	0.6769	0.5146

Table 3. NIQE evaluation results of different algorithms

View table

Table 3. NIQE evaluation results of different algorithms

Algorithm	LOL	DICM	LIME	NPE	Average
Dong	12.4560	9.0300	11.9810	15.2693	12.1841
KinD++	10.1830	10.4091	14.2089	14.8595	12.4151
RetinexNet	12.3950	8.6770	10.8850	12.1360	11.0233
Retinexformer	12.2430	10.8970	14.7000	15.9380	13.4445
EnlightGAN	10.2600	10.3330	14.7780	16.4783	12.9623
Zero-DCE	12.9740	13.4240	11.9870	15.6460	13.5078
Proposed algorithm	11.9784	8.3311	10.8381	13.7482	11.2240

Table 4. BRISQUE evaluation results of different algorithms

View table

Table 4. BRISQUE evaluation results of different algorithms

Algorithm	LOL	DICM	LIME	NPE	Average
Dong	0.4964	0.5400	0.6397	0.6309	0.5768
KinD++	0.5030	0.4871	0.4487	0.4559	0.4737
RetinexNet	0.5040	0.4960	0.5460	0.5660	0.5280
Retinexformer	0.5050	0.5150	0.4640	0.5310	0.5038
EnlightGAN	0.4990	0.5260	0.4920	0.4939	0.5027
Zero-DCE	0.4960	0.5460	0.4970	0.4870	0.5065
Proposed algorithm	0.4883	0.5062	0.4610	0.4601	0.4789

Table 5. Runtime of different images processed by different algorithms

View table

Table 5. Runtime of different images processed by different algorithms

Algorithm	Running time /s
Algorithm	LOL	DICM	LIME	NPE	Average
Dong	0.4467	0.6909	1.2853	1.4615	0.9711
KinD++	4.0399	6.6172	5.4900	4.2246	5.0929
RetinexNet	0.6213	0.9614	0.8773	0.6852	0.7863
Retinexformer	0.2445	0.3428	0.4659	0.4420	0.3738
EnlightGAN	0.2144	0.2835	0.2586	0.2192	0.2439
Zero-DCE	0.03317	0.0692	0.2201	0.2624	0.1462
Proposed algorithm	0.8475	1.0528	1.1242	0.7887	0.9533

Table 6. Comparison of enhancement effects of different modules on low-light images
View table
Table 6. Comparison of enhancement effects of different modules on low-light images
ABA BACE AIF PSNR SSIM NIQE BRISQUE
√ × × 12.8264 0.4523 13.1397 0.5844
√ √ × 16.0354 0.4963 12.1509 0.4789
√ × √ 15.4667 0.5306 13.0009 0.5868
√ √ √ 19.0364 0.5869 12.1357 0.4681

Tools

Get Citation

Copy Citation Text

Chengkang Yu, Guangliang Han. HSV Space-Based Nonlinear Adaptive Low-Light Image Enhancement Algorithm[J]. Laser & Optoelectronics Progress, 2025, 62(8): 0837001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites