SuPeR: Milano Retinex implementation exploiting a regular image grid

Michela Lecca; Stefano Messelodi

doi:10.1364/JOSAA.36.001423

Journal of the Optical Society of America A
Vol. 36,
Issue 8,
pp. 1423-1432
(2019)
•https://doi.org/10.1364/JOSAA.36.001423

SuPeR: Milano Retinex implementation exploiting a regular image grid

Michela Lecca and Stefano Messelodi

Not Accessible

Your library or personal account may give you access

Get PDF
Email
Share
Get Citation
Copy Citation Text
Michela Lecca and Stefano Messelodi, "SuPeR: Milano Retinex implementation exploiting a regular image grid," J. Opt. Soc. Am. A 36, 1423-1432 (2019)

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

Check for updates

Abstract

A spatial color algorithm grounded on the Retinex theory is known as a Milano Retinex. This type of algorithm performs image enhancement by processing spatial and color cues in the neighborhood of each image pixel. Because this local, pixel-wise analysis is time consuming, optimization techniques are needed to expand the use of Milano Retinexes to applications that require fast or even real-time image processing. In this work, we propose SuPeR, an efficient optimization of the Milano Retinex local spatial color processing that exploits superpixels, which are as the regular, rectangular blocks of a grid that partitions the image support. Image enhancement is obtained by reworking channel-wise the intensity of each pixel based on the maximum color intensities of the blocks and on its distance from the blocks. The experiments, carried out on real-world image datasets, show good performance compared to other Milano Retinexes.

Full Article | PDF Article

More Like This

Generalized equation for real-world image enhancement by Milano Retinex family

Michela Lecca
J. Opt. Soc. Am. A 37(5) 849-858 (2020)

GREAT: a gradient-based color-sampling scheme for Retinex

Michela Lecca, Alessandro Rizzi, and Raul Paolo Serapioni
J. Opt. Soc. Am. A 34(4) 513-522 (2017)

Tuning the locality of filtering with a spatially weighted implementation of random spray Retinex

Michela Lecca and Alessandro Rizzi
J. Opt. Soc. Am. A 32(10) 1876-1887 (2015)

Previous Article Next Article

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 (8)

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 (4)

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 (14)

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	$f_{0}$	$f_{1}$	$f_{2}$ [ $\times 10^{- 3}$ ]	$Δ E$	NIQE	BRISQUE
INPUT	64.30	15.34	4.13	–	7.01	26.28
SuPeR ( $M = 25$ )	105.80	21.13	2.98	25.65	7.37	25.41
SuPeR ( $M = 100$ )	119.06	21.30	2.93	32.76	7.44	27.05
SuPeR ( $M = 400$ )	128.26	20.92	3.05	37.39	7.59	26.74
SuPeR ( $M = 900$ )	132.23	20.59	3.13	39.32	7.62	26.19
SuPeR ( $M = 1600$ )	134.52	20.34	3.19	40.42	7.74	25.59
STAR ( $\bar{K} = 160$ ) [26]	98.62	19.98	3.16	20.82	7.14	26.43
GREAT [25,26]	96.93	19.74	3.29	18.88	7.19	24.87
GRASS [23,26]	89.19	20.81	3.14	16.65	6.82	22.59
RSR ( $N = 20$ , $m = 150$ ) [16,26]	87.41	20.66	3.20	14.33	7.16	23.63
L-ETR [15,26]	104.30	23.13	2.67	25.15	7.24	24.29
HE-C [26]	131.62	25.65	1.55	43.99	7.79	26.91
HE-L [26]	129.96	26.04	2.04	38.94	7.81	28.51
CLAHE-C ( $X = 4$ ) [26]	100.72	27.09	2.23	27.86	7.81	27.68
CLAHE-L ( $X = 4$ ) [26]	93.61	26.31	2.45	23.49	7.79	26.91
NPEA [36,26]	99.60	20.49	3.04	24.56	7.33	28.36
LIME [35,26]	125.81	27.81	2.21	32.92	8.26	30.24
NL-RETINEX [37,26]	84.18	22.32	3.08	23.34	7.49	26.86
FireFly [38]	104.23	19.85	2.95	26.48	8.85	30.37
ACE [39]	126.51	20.39	2.97	43.05	7.00	26.69
SLRMSR [20]	101.36	22.41	2.82	26.91	7.37	27.11

Algorithm	$f_{0}$	$f_{1}$	$f_{2}$ [ $\times 10^{- 3}$ ]	$Δ E$	NIQE	BRISQUE
FLIR20-Le (none)	32.39	6.67	6.00	–	3.79	39.23
SuPeR ( $M = 25$ )	110.40	13.89	4.38	46.98	4.28	39.38
SuPeR ( $M = 100$ )	131.84	13.24	4.55	55.77	4.29	38.31
SuPeR ( $M = 400$ )	146.38	12.28	4.79	61.06	4.35	38.08
SuPeR ( $M = 900$ )	152.07	11.59	4.98	63.00	4.42	36.63
SuPeR ( $M = 1600$ )	155.17	11.27	5.06	64.02	4.43	37.14
STAR ( $\bar{M} = 1538$ ) [26]	124.54	12.43	4.73	52.13	4.28	38.93
GREAT [25,26]	106.71	12.59	4.71	43.18	4.23	41.12
RSR ( $N = 20$ , $m = 200$ ) [16,26]	87.00	15.09	4.06	34.93	4.46	25.93
HE-C [26]	139.14	24.65	3.95	57.45	4.48	40.92
HE-L [26]	138.37	24.76	4.72	54.89	4.45	40.89
CLAHE-C ( $X = 8$ ) [26]	78.20	20.38	3.35	32.02	3.85	35.67
CLAHE-L ( $X = 8$ ) [26]	69.98	18.71	3.73	25.97	3.78	38.12
NPEA [36,26]	118.84	15.89	4.26	51.50	4.46	42.15
LIME [35,26]	113.18	17.18	3.75	47.51	4.98	46.01
NL-RETINEX [37,26]	68.56	18.90	3.57	28.29	4.17	32.37
FireFly [38]	93.01	11.36	4.76	39.54	4.11	41.07
ACE [39]	127.41	11.60	4.68	55.83	4.02	40.53
SLRMSR [20]	122.43	14.99	4.26	52.70	3.90	36.04

Dataset	$N_{k p}$	$N_{t m}$	$N_{c m}$	$\frac{N_{c m}}{N_{t m}}$
FLIR20-He	588	–	–	–
FLIR20-Le	117	68	41	0.60
SuPeR ( $M = 25$ )	1046	207	189	0.91
STAR [26]	680	154	140	0.91
GREAT [25,26]	625	166	152	0.92
RSR ( $N = 20$ , $m = 200$ ) [16,26]	807	232	214	0.92
HE-C [26]	2660	385	165	0.43
HE-L [26]	2619	369	155	0.42
CLAHE-C [26]	2227	436	201	0.46
CLAHE-L ( $X = 8$ ) [26]	1670	416	197	0.47
NPEA [36,26]	2233	391	113	0.29
LIME [35,26]	1260	444	270	0.61
NL-RETINEX [37,26]	1367	364	161	0.44
FireFly [38]	475	176	162	0.92
ACE [39]	500	211	201	0.94
SLRMSR [20]	1653	192	169	0.88

N. Blocks	25	100	400	900	1600
Time [s]	0.18	0.23	0.26	0.31	0.34

Algorithm	$f_{0}$	$f_{1}$	$f_{2}$ [ $\times 10^{- 3}$ ]	$Δ E$	NIQE	BRISQUE
INPUT	64.30	15.34	4.13	–	7.01	26.28
SuPeR ( $M = 25$ )	105.80	21.13	2.98	25.65	7.37	25.41
SuPeR ( $M = 100$ )	119.06	21.30	2.93	32.76	7.44	27.05
SuPeR ( $M = 400$ )	128.26	20.92	3.05	37.39	7.59	26.74
SuPeR ( $M = 900$ )	132.23	20.59	3.13	39.32	7.62	26.19
SuPeR ( $M = 1600$ )	134.52	20.34	3.19	40.42	7.74	25.59
STAR ( $\bar{K} = 160$ ) [26]	98.62	19.98	3.16	20.82	7.14	26.43
GREAT [25,26]	96.93	19.74	3.29	18.88	7.19	24.87
GRASS [23,26]	89.19	20.81	3.14	16.65	6.82	22.59
RSR ( $N = 20$ , $m = 150$ ) [16,26]	87.41	20.66	3.20	14.33	7.16	23.63
L-ETR [15,26]	104.30	23.13	2.67	25.15	7.24	24.29
HE-C [26]	131.62	25.65	1.55	43.99	7.79	26.91
HE-L [26]	129.96	26.04	2.04	38.94	7.81	28.51
CLAHE-C ( $X = 4$ ) [26]	100.72	27.09	2.23	27.86	7.81	27.68
CLAHE-L ( $X = 4$ ) [26]	93.61	26.31	2.45	23.49	7.79	26.91
NPEA [36,26]	99.60	20.49	3.04	24.56	7.33	28.36
LIME [35,26]	125.81	27.81	2.21	32.92	8.26	30.24
NL-RETINEX [37,26]	84.18	22.32	3.08	23.34	7.49	26.86
FireFly [38]	104.23	19.85	2.95	26.48	8.85	30.37
ACE [39]	126.51	20.39	2.97	43.05	7.00	26.69
SLRMSR [20]	101.36	22.41	2.82	26.91	7.37	27.11

Abstract

Cited By

Figures (8)

Tables (4)

Equations (14)

Journal of the Optical Society of America A