Color - Vision with Direction - page 24

( y + b ), where y is a shorthand way of saying “red plus green”.

The latter is perceived as yellow light if the amount of light in the red wavelength
range is approximately the same as that of green. Together, the ( r + g

− ) and ( b + y −

− )-, ( g + r − )-,

)-cells populate the vast majority of the cells in layers 3-6 of
LGN. Albeit in minority, there is another significant cell type in these layers that is of
the center-surround type. Cells of this type differ from the other cells in that they are
color-insensitive, and presumably implement the L + M -scheme. Additionally, the
entire layers 1 layers 2 are populated by this type of cells, the magnocellular cells ,
albeit these are larger than the parvocellular cells populating layers 3-6.
It is worth noting that in the perception of lightness, or luminosity, the blue color
does not play a significant role. Details only differing in the amount of blue do not
show up very well because such changes do not contribute to the perception of edges
and lines.
In LGN, most of neurons are wavelength-selective while being center-surround .
They are excited by one wavelength pattern of the stimulus light falling in one re-
gion of their receptive field and inhibited by another in the other. However, they do
not measure wavelength differences between the light falling into their center, and
surround. Merely, they express the difference in the amount of light quanta with
specific wavelengths captured by the center and surround regions. In a way, it is a
matter of spatial subtraction that these cells perform, not wavelength subtraction.
The blobs encountered in layers 1-3 of the V1 area contain the so-called double-
opponent color cells , which are sensitive to wavelength differences in the center and
surround regions [155]. They respond vigorously to one wavelength in the center
of their receptive field, while they are inhibited by another (still in the center). The
same cells are excited by this second wavelength in the surround and depressed by
the first. A double-opponent color cell can thus be excited by the wavelength of red
and inhibited by that of green in its center, while it will be excited by the wavelength
of green and inhibited by the wavelength of red in the surround. This behavior has
been denoted as ( r + g

( y + b − )-, and ( b + y −

) in neurobioligical studies. Consequently, a large
patch reflecting red will generate zero response from these cells because the wave-
length pattern in the center is “subtracted” from that of the surround. In fact, not only
red colored light, but any colored light, including white, that shines up a large patch
observed by an ( r + g

− /g + r −

)-cell
of LGN, by contrast, will be excited, if the wavelength pattern matches either the
one it prefers in the center or the one in the surround. The following types of double-
opponent color cells have been experimentally observed in blobs: ( r + g

− /g + r − )-cell will generate zero response. An ( r + g −

− /r − g +),

( r b +), where b corresponds

to the light with wavelength patterns of S-cones (blue) and y is light with an addi-
tive combination of wavelength patterns represented by L-cones (red) and M-cones
(green). It is presumably the double-opponent color cells that are largely responsible
for color constancy observed in many fish species, macaque monkey and humans,
although these cells appear in the retina in fish.
Simplified, there are three color axes (lightness, red-green and yellow-blue)
along which color processing takes place in humans. However, there are only three
independent measurements, represented by the signals L , M , and S , that drive our

− g + /r + g − ), ( b + y − /y + b − ), ( b − y + /y −