This algorithm is based on the idea that the human eye has the feature
of color constancy which ensures that the perceived color of objects
remains relatively constant under varying illumination conditions.
The effect was described in 1971 by Edwin Land, who formulated retinex
theory to explain it.
Wikipedia: http://en.wikipedia.org/wiki/Retinex

And this algorithm is an approximation to convert images that contain
shadows and take an approximation of what the human eye was viewing
when took the photo.

I put the links of some screenshots and the introduction of the paper:


On 4/22/06, Nathan Summers <[EMAIL PROTECTED]> wrote:
> On 4/21/06, Pedro Alonso <[EMAIL PROTECTED]> wrote:
> > Hello,
> >
> > I would like to develop a plug-in for The Gimp, but I don't know if it
> > would be accepted as Google SoC project or need to be a contribution
> > to the core of the program.
> Depending on the nature of the plugin, it could very well be accepted
> as part or all of a SoC project.
> > If it were possible my purpose is:
> > - Implements the algorithm based on the paper "capturing a black cat
> > in shade:                      past and present of Retinex color
> > appearance models"
> That paper doesn't appear to be online.  Could you provide a summary
> of the algorithm and its effects?
> Rockwalrus
Capturing a black cat in shade: past and present
      of Retinex color appearance models
                     John J. McCann
                     McCann Imaging
                Belmont, Massachusetts 02478
                  E-mail: [EMAIL PROTECTED]

This work recounts the research on capturing real-life
scenes, calculating appearances, and rendering sensations on film
and other limited dynamic-range media. It describes the first pat-
ents, a hardware display used in Land’s Ives Medal Address in
1968, the first computer simulations using 20 24 pixel arrays, psy-
chophysical experiments and computational models of color con-
stancy and dynamic range compression, and the Frankle-McCann
computationally efficient retinex image processing of 512 512 im-
ages. It includes several modifications of the original approach, in-
cluding recent models of human vision and gamut-mapping applica-
tions. This work emphasizes the need for parallel studies of
psychophysical measurements of human vision and computational
algorithms used in commercial imaging systems. © 2004 SPIE and
IS&T. [DOI: 10.1117/1.1635831]

1 Introduction

In Land’s first lecture at a Friday Evening Discourse at the
Royal Institution, London, on 28 April 1961, he used real
papers as a part of a series of experiments including red and
white projections.1 For Land, it was the turning point from
photographic projections to experiments with controlled re-
flectance and illuminants. More important, it was the turn-
ing point from the dimensionless coordinate system as a
physical description of the stimulus to the psychophysical
quantity lightness as the determinant of color. Up until this
lecture, Land had been trying to correlate the colors he saw
with the physical stimulus. He knew that colorimetry was
of little help beyond calculating quanta catch of receptors.
His experiments with Daw showed that adaptation, specifi-
cally the change of receptor sensitivity in response to light,
could not account for color appearance. They projected red
and white images of ambiguous objects to naive observers
using 1- s duration flashes. Color memory and adaptation
could not explain the colors in ambiguous displays seen for
the first time with so few photons.2 Land knew spatial fac-
tors were important, but he did not know how to put the
model of human color vision together. In his process of
persistent exploration, he made the critical observation that
color appearance correlated with the triplet of lightness ap-
pearances in long L -, middle M -, and short S -wave
light.3,4 This idea created a halfway point between the
physical measurement of cone quanta catch and color ap-
pearance. If we found a physical model whose output cor-
related the appearances ranging from white to black, then
that mechanism could be used three times in parallel to
predict colors. This observation transformed the study of
color to a need for understanding how the eye sees whites,
grays, and blacks.
    Land’s observation still stands. The triplet of apparent
lightnesses correlates with color. The observation is impor-
tant because a variety of different phenomena can influence
lightness, such as simultaneous contrast, the Cornsweet ef-
fect, assimilation, and spatial blur of the retinal image. Re-
gardless of the cause of the lightness shifts, when two iden-
tical physical objects look different, color appearances
correlate with their L, M, and S lightnesses.5,6 In an effec-
tive color assimilation display, there are two sets of nine
square red-brown patches on a yellow and blue striped
background. On the left, the red-brown patches fall on top
of the yellow stripes, and on the right they fall on the blue
stripes. The left patches appear a purple red, while the right
patches appear a yellow orange. In other words, the left
patches appear more blue and the right ones more yellow.
Color assimilation displays exhibit larger color effects than
color contrast.7 In assimilation, predominantly black sur-
rounds make grays appear darker, while in contrast, black
surrounds make grays appear lighter. Figure 1 shows the
color display and the R, G, B separations for this effective
color assimilation display. Identical square patches appear
different colors. In the R separation, the corresponding
patches are lighter on the right; in the G separation, the
patches on the right are lighter; and in the B separation, the
patches are darker on the right. Whenever R and G separa-
tions are lighter and B separation is darker, then that patch
will appear more yellow. Whenever B separation is lighter
and R and G separations are darker, then that patch will
appear more blue. Colors correlate with R, G, B
lightnesses.6 This is the theory that Land proposed 40 years
ago and called Retinex,3 and was the basis for the sympo-
sium ‘‘Retinex at 40.’’

Gimp-developer mailing list

Reply via email to