Performance comparison of image enhancers with and without deep learning

Michela Lecca; Fabio Poiesi

doi:10.1364/JOSAA.446969

Journal of the Optical Society of America A
Vol. 39,
Issue 4,
pp. 610-620
(2022)
•https://doi.org/10.1364/JOSAA.446969

Performance comparison of image enhancers with and without deep learning

Michela Lecca and Fabio Poiesi

Not Accessible

Your library or personal account may give you access

Get PDF
Email
Share
Get Citation
Copy Citation Text
Michela Lecca and Fabio Poiesi, "Performance comparison of image enhancers with and without deep learning," J. Opt. Soc. Am. A 39, 610-620 (2022)

Export Citation
- BibTex
- Endnote (RIS)
- HTML
- Plain Text
Citation alert
Save article

Check for updates

Abstract

Image enhancement is a computational procedure to improve visibility of details and content of an input image. Several image enhancement algorithms have been developed thus far, from traditional methods that process a single image based on physical models of image acquisition and formation to recent deep learning techniques, where enhancement models are learned from data. Here, we empirically compare a set of traditional and deep learning enhancers, which we selected as representing different methodologies for the improvement of poorly illuminated images. Our experiments are conducted on public data and show that, although all the considered enhancers improve the visibility of the image content and details, the deep-learning-based methods generally produce less noisy images than the traditional ones. This last outcome has to be carefully considered when enhancers are used as preprocessing for algorithms that are sensitive to noise. As a case study, and with the purpose to promote more aware usage of these two groups of enhancers in computer vision applications, we discuss the impact of image enhancement in the framework of image retrieval performed through two popular algorithms, i.e., SIFT and ORB, implementing different image descriptions and having different sensitivities to noise.

Full Article | PDF Article

More Like This

Deep image enhancement for ill light imaging

Rizwan Khan, You Yang, Qiong Liu, Jialie Shen, and Bing Li
J. Opt. Soc. Am. A 38(6) 827-839 (2021)

Image dehazing combining polarization properties and deep learning

Ke Suo, Yaowen Lv, Jiachao Yin, Yang Yang, and Xi Huang
J. Opt. Soc. Am. A 41(2) 311-322 (2024)

CODEN: combined optimization-based decomposition and learning-based enhancement network for Retinex-based brightness and contrast enhancement

Sangjae Ahn, Joongchol Shin, Heunseung Lim, Jaehee Lee, and Joonki Paik
Opt. Express 30(13) 23608-23621 (2022)

Previous Article Next Article

Data availability

Data underlying the results presented in this paper are available in [28,29].

28. C. Wei, W. Wang, W. Yang, and J. Liu, “Lol dataset,” 2011, https://daooshee.github.io/BMVC2018website/.

29. TeV Resources, “Mexico-2020,” available at https://tev.fbk.eu/resources/imageenhancement.

Cited By

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.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Figures (5)

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.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Tables (5)

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.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Equations (2)

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.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Algorithm	$B$	$C$	$f [\times 10^{- 3}]$	N
Ground-Truth	114.58	14.18	3.52	4.25
Input	14.82	3.10	6.67	5.72
SuPeR [7]	110.14	19.69	2.85	8.86
Light-RSR [8]	153.05	29.97	2.01	9.85
LIME [9]	87.95	19.14	2.99	8.79
FUSION [11]	84.17	15.29	3.73	9.71
NPEA [10]	94.46	17.20	3.39	9.14
RetinexNet [6]	107.43	14.33	3.99	9.73
MBLLEN [12]	98.56	16.99	2.91	3.27
EnlightenGAN [13]	89.47	14.54	3.66	4.82
MiRNet (LOL) [14]	115.65	14.68	3.35	3.48
MiRNet (FIVEK) [14]	43.41	10.12	4.77	3.83
ERL [15]	29.94	5.92	5.86	6.75

Algorithm	$B$	$C$	$f [\times 10^{- 3}]$	N
Ground-Truth	115.73	14.85	3.29	4.11
Input	14.91	3.20	6.59	6.04
SuPeR [7]	89.93	17.45	3.14	9.14
Light-RSR [8]	150.27	30.04	1.89	10.09
LIME [9]	85.43	18.89	3.13	9.16
FUSION [11]	81.65	15.16	3.77	9.94
NPEA [10]	93.95	17.47	3.42	9.38
RetinexNet [6]	103.87	14.24	3.96	10.22
MBLLEN [12]	91.52	16.20	2.98	3.51
EnlightenGAN [13]	87.81	14.14	3.75	4.92
MiRNet (LOL) [14]	114.94	14.60	3.30	3.46
MiRNet (FIVEK) [14]	39.77	9.67	4.86	3.98
ERL [15]	28.43	6.03	5.81	7.06

T [s]	$B$	$C$	$f [\times 10^{- 3}]$	$N$	$K p^{S}$	$g m^{S}$	$b m^{S}$	$M^{S}$	$M T^{S}$	$K p^{O}$	$g m^{O}$	$b m^{O}$	$M^{O}$	$M T^{O}$
6	21.28	3.44	6.60	5.08	35	6	3	56.00	40.00	58	12	10	60.00	54.00
15	34.28	6.26	5.84	4.64	137	66	12	88.00	88.00	201	98	33	88.00	88.00
27	80.95	15.37	3.76	3.83	741	–	–	–	–	457	–	–	–	–

	T [s]	$B$	$C$	$f [\times 10^{- 3}]$	$N$	$K p^{S}$	$g m^{S}$	$b m^{S}$	$M^{S}$	$M T^{S}$	$K p^{O}$	$g m^{O}$	$b m^{O}$	$M^{O}$	$M T^{O}$
SuPeR	6	122.64	14.07	4.27	4.57	1618	279	127	100.00	68.00	490	170	93	100.00	100.00
	15	115.68	15.34	3.79	3.97	1225	425	107	100.00	98.00	491	241	91	100.00	100.00
	27	118.32	18.81	3.00	3.33	1139	–	–	–	–	499	–	–	–	–
Light-RSR	6	142.70	17.76	3.72	4.71	2485	377	190	100.00	70.00	493	195	102	100.00	98.00
	15	131.62	19.07	3.37	4.11	1819	564	153	100.00	98.00	496	268	92	100.00	100.00
	27	128.71	21.50	2.88	3.56	1396	–	–	–	–	497	–	–	–	–
LIME	6	104.99	14.48	4.27	4.61	1221	284	142	100.00	90.00	489	126	111	100.00	100.00
	15	116.52	17.69	3.60	4.10	1405	512	145	100.00	100.00	495	198	115	100.00	100.00
	27	144.51	21.63	2.91	3.57	1620	–	–	–	–	498	–	–	–	–
FUSION	6	109.87	9.92	5.23	5.14	966	171	125	100.00	70.00	453	114	102	100.00	96.00
	15	109.62	11.58	4.80	4.37	1021	346	128	100.00	100.00	469	190	108	100.00	100.00
	27	124.85	15.94	3.81	3.57	1292	–	–	–	–	490	–	–	–	–
NPEA	6	119.51	16.44	4.23	4.75	3722	252	236	100.00	34.00	498	73	121	100.00	96.00
	15	113.52	15.20	4.25	4.06	2249	401	183	100.00	70.00	497	148	133	100.00	100.00
	27	116.10	16.33	3.73	3.41	1288	–	–	–	–	490	–	–	–	–

	T [s]	$B$	$C$	$f [\times 10^{- 3}]$	$N$	$K p^{S}$	$g m^{S}$	$b m^{S}$	$M^{S}$	$M T^{S}$	$K p^{O}$	$g m^{O}$	$b m^{O}$	$M^{O}$	$M T^{O}$
RetinexNet	6	135.42	8.17	5.75	6.14	1331	79	243	98.00	34.00	424	25	131	92.00	50.00
	15	138.08	9.55	5.44	5.57	1836	132	433	100.00	70.00	464	19	169	100.00	40.00
	27	147.08	12.51	4.68	4.51	1970	–	–	–	–	493	–	–	–	–
MBLLEN	6	59.96	14.61	3.76	3.77	445	208	38	100.00	100.00	434	203	91	100.00	100.00
	15	64.47	15.76	3.52	3.86	547	324	39	100.00	100.00	461	277	77	100.00	100.00
	27	84.48	20.05	2.77	3.85	864	–	–	–	–	484	–	–	–	–
Enlighten GAN	6	127.58	12.48	4.50	4.15	972	250	116	100.00	94.00	481	125	138	100.00	100.00
	15	139.41	14.75	4.03	3.71	1191	412	134	100.00	98.00	496	165	153	100.00	100.00
	27	148.99	17.43	3.48	3.28	1268	–	–	–	–	496	–	–	–	–
MIRNet (LOL)	6	69.36	9.42	4.69	4.14	384	88	55	94.00	90.00	380	77	108	98.00	86.00
	15	93.22	12.12	3.87	3.57	626	217	77	100.00	100.00	463	154	122	100.00	98.00
	27	133.62	17.34	2.82	3.04	988	–	–	–	–	494	–	–	–	–
MIRNet (FIVEK)	6	85.47	8.95	5.31	3.86	473	118	59	100.00	92.00	352	85	78	100.00	96.00
	15	101.53	11.21	4.70	3.45	975	343	110	100.00	100.00	469	173	103	100.00	100.00
	27	124.19	15.88	3.62	3.10	1436	–	–	–	–	499	–	–	–	–
ERL	6	39.61	5.12	5.99	5.06	113	35	13	88.00	82.00	160	56	32	88.00	82.00
	15	52.89	9.10	4.82	4.58	467	229	41	98.00	100.00	352	175	65	100.00	98.00
	27	91.37	16.56	3.02	3.78	1042	–	–	–	–	496	–	–	–	–

Abstract

Data availability

Cited By

Figures (5)

Tables (5)

Equations (2)

Journal of the Optical Society of America A