Land and McCann Retinex
Review papers on L&M Retinex (1969-2021)
1969
Rod vision is physiologically distinct from cone vision. It captures light and generates an achromatic lightness images. When the spectra of rod stimuli and long-wave cone stimuli are balanced, the rods act as a short-wave color channel. Rods are a perfectly good color receptors when stimulated properly.
Rods color appearances provide evidence of independent spatial channels that control color sensations. Rods color appearances support Retinex Theory.
1969
J.J. McCann, and J. L. Benton
" Interactions of the Long-Wave Cones and the Rods to Produce Color Sensations",
J. opt. Soc. Am., 59, 103-107, 1969.
_______________
1970
Edwin Land’s Ives Medal Address was in 1967. The JOSA paper was not published until 1971. In 1970 McCann, Land, and Tatnall described a technique for quantifying lightness to be used as ground truth in computational models. It emphasized that sensations, not physical reflectance was the purpose of the ground truth measurements, and that the goal of the Lightness model was to calculate sensations.
“Lightness1, 2, 3, 4 is the family of sensations from white to black that a person
sees. Lightness is the output of a biological system. It is a sensation. There is
no physical definition for lightness because it is not necessarily related to a
physical quantity of light from a point. either in radiometric terms or photometric
terms.”
“Although these test situations do not necessarily provide a complete general
test of any model, a particular model that can predict lightnesses in all
these situations is certainly of great value. We propose to study all of these
situations using the Standard Lightness Display described here to quantify both
observers' sensations and model's predictions on the same equally spaced scale.
Thus, using the techniques described in this paper. we propose to test our present
ideas, shape our future models, and find, if possible. a single general model
that can predict the lightnesses in all the different situations discussed.”
1970
J. J. McCann, E. H. Land and S. M. V. Tatnall,
“Technique for Comparing Human Visual Responses with a Mathematical Model for Lightness‚”
Am. J. Optometry and Archives of Am. Acad. Optometry, 47(11), 845-855, 1970.
_______________
1974
Land wrote a paper that expanded the Retinex model for calculating Lightness. This 1974 publication documented Land’s Friday Evening Discourse at the Royal Society, London. The text expanded the idea that lightness was a sensation that correlated with reflectance much of the time, but not all the time. This paper emphasized that lightness was the “biological correlate” of reflectance. It emphasized the color matching as the method of quantifying lightness and color sensations.
1974
E. H. Land,
“The Retinex Theory of Colour Vision”,
Proc. R. Instn Gr Brit. 47. 23-58, (1974).
_______________
1976
The McCann, McKee and Taylor paper described three results. First, the quantitative measurements of color sensations for ground truth data for five different uniform Mondrian illuminations. The data matches showed sensations have a strong, correlation with physical spectral reflectance with systematic limitations. The departures from perfect correlation were shown to be caused by the overlap in spectral sensitivity of cone visual pigments.
Second, The paper described the details of a Fortran computer program that used as input radiances measured with human cone spectral sensitivities. The model’s input included crosstalk’s spectral limitations. Further, it described the calculation details of the first Retinex computer model that used a series of paths. The length of the paths was tuned to optimize the models fit to ground truth data.
Third, it showed that “computed lightness” correlates with “observed lightness”. Ratios of cone quanta catch that included crosstalk can predict observer color sensation matches.
The paper did not report that observer matched the same reflectance chip in the Munsell Book in uniform Illumination. The model did not calculate physical reflectance. Rather it calculated “Scaled Integrated Reflectance” that includes cone crosstalk that results from spatial comparisons..
1976
J. McCann, S.P. McKee & T. Taylor,
"Quantitative Studies in Retinex Theory, A Comparison BetweenTheoretical Predictions and Observer Responses to Color Mondrian Experiments",
Vision Res., 16, 445-58, (1976).
_______________
1983
Frankle and McCann replaced paths with multiresolution spatial comparisons. Lightnesses were calculated using the same digital operations: ratio, threshold, product, reset, and average. The hardware used to calculate sensations had enough digital memory to hold the array of 512 by 512, by 24 bit RGB image. The system was programmed to perform Retinex operations on the entire array in small fractions of a second. The combination of multiresolution and hardware image processing made the calculation extremely efficient.
1983
Efficient Multi-resoluton Computations
-
J.Frankle and J. McCann,
“Method and apparatus of lightness imaging”,
US Patent, 4,384,336, May 17, 1983.
_______________
1999
The paper reviews the advances in Land and McCann Retinex since 1971. It summarized the extended study of the visibility of gradients and cone crosstalk. It emphasized that L&M Retinex is based on gradients and edges, rather than reflectances and illumination. Using Maximov’s Shoebox, it demonstrated that complex arrays of different reflectances have constant matching appearances. It introduced color gamut mapping using spatial comparisons.
1999
J. McCann,
"Lessons learned from Mondrians applied to real images and color gamuts," Seventh Color Imaging Conference: Color Science, Systems and Applications 1-8, (1999).
_______________
2004
At the Retinex at Forty Symposium at the Electronic Imaging Conference presented papers on the developments of Land and McCann Retinex. Among other topics were an annotated bibliography of Red and White photography, and a review of past and present computational models of lightness.
2004
J. McCann, editor,
“Retinex at 40 Symposium‚”Journal Special Collection,
JEI, 13, 1-145, 2004.
J. McCann,
"Guest Editorial: Special Section on Retinex at 40," Journal of Electronic Imaging 13(1), 6-7 (2004)
J. McCann, J. L. Benton and S. P. McKee,
"Red/white projections and rod/long-wave cone color: an annotated bibliography," Journal of Electronic Imaging 13(1), 8-14 (2004).
J. McCann,
"Capturing a black cat in shade: past and present of Retinex color appearance models," Journal of Electronic Imaging 13(1), 36-47 (2004).
B. Funt, and F. Ciurea, "Retinex in MATLAB (TM)," Journal of Electronic Imaging 13(1), 48-57 (2004).
_______________
2004-2005
Observer matches measured the departures from perfect color constancy using 27 different illuminants. The data showed that cone crosstalk, not adaptation, correlates with appearance.
2004
J.McCann,
"Mechanism of color constancy," 12th Color Imaging Conference, 12, 29-36 (2004).
2005
J. McCann,
"Do humans discount the illuminant?," in Human Vision and Electronic Imaging X, pp. 9-16, SPIE, San Jose, CA, USA, SPIE Proc, vol. 5666, pp. 9-16, (2005).
_______________
2012
Chapter 32 in The Art and Science of HDR Imaging provides the most comprehensive review of Retinex algorithms form 1967 through 2012. It attempts to clarify common misconceptions, and identify critical difference among various Retinexes.
2012
J. McCann, A. Rizzi,
“The Art and Science of HDR Imaging‚”,
IS&T Wiley, Chichester (2012).
2014
_______________
Observers measured the appearances of facets in two identical 3-D Mondrians reflectance targets. One was illuminated in nearly uniform illumination (LDR); the other in highly-directional illumination (HDR) with sharp shadows. Traditional flat Mondrians in uniform illumination show high correlation between sensations and Scaled Integrated Reflectance. 3-D Mondrians in nearly uniform LDR illumination shows that some facets have poor correlation with reflectance. In HDR illumination, with edges in both reflectance and illumination, there are many facets that show little or no correlation. Appearance is controlled by the scene’s edges and gradient, regardless of whether they are from reflectance or illumination.
2014
J.J. McCann , C. Parraman, & A. Rizzi,
“Reflectance, illumination, and appearance in color constancy‚”
Frontiers in Psychology, 24, January 2014, 00005
<http://dx.doi.org/10.3389/fpsyg.2014.00005>
_______________
2017
Today, there are many different Retinex algorithms. This special section, “Retinex at 50,” describes a wide variety of them, along with their different goals, and ground truths used to measure their success. This paper reviews (and provides links to) the original Retinex experiments and image-processing implementations. Observer matches (measuring appearances) have extended our understanding of how human spatial vision works. This paper describes a collection very challenging datasets, accumulated by Land and McCann, for testing algorithms that predict appearance.
2017
J. McCann
"Retinex at 50: color theory and spatial algorithms, a review,"
-
J.Electron. Imaging 26(3), 031204 (2017), doi: 10.1117/1.JEI.26.3.031204.
http://dx.doi.org/10.1117/1.JEI.26.3.031204.
_______________
Meters can measure scene radiances, but do not provide reliable information about images on retinas. Retinal image contrast is substantially reduced compared to scene radiances. Intaocular scatter, namely light from all part of the field of view, transforms scene information. Retinal contrast, not scene radiance is the appropriate input for computational models of neural image processing in vision.
This paper provides a Matlab program to calculate the image on the retina. Further, it shows that the scene content change the post-receptor neural processing.
2017
J. McCann and Vassilios Vonikakis,
“Calculating Retinal Contrast from Scene Content: A Program”,
Front. Psychol. | doi: 10.3389/fpsyg.2017.02079 (2017).
https://www.frontiersin.org/articles/10.3389/fpsyg.2017.02079/full
Historical flowchart of the literature (with links) surrounding
Land and McCann’s “Lightness and Retinex Theory”
J. Opt. Soc. Am., 61, pp. 1-11, 1971