o L+M: this and above are good for detecting light
We don’t see color, we see wavelengths. The more o (L+M)-S: S does very little.
wavelengths absorbed, the darker the colour we see. 3. Appearance: perceived colors should be assigned to
The color of what we see depends on which the object the wavelengths came from, and should
wavelengths get reflect and hit the eye. 3 Steps: stay the same under different lighting conditions.
1. Detection: wavelengths must be detected.
With the 3-number system, we can see 2 million
We have 3 cone photoreceptors: S-cones (short, color variations (or 26,000 colors). Color space: it’s
blue; 420), M-Cones (medium, green; 535) and L- 3D and established because color perception is
cones (long, red; 565). based on outputs of 3 cone types. Achromatic: any
S-cones are rarer and less sensitive color that lacks hue (chromatic component): black
2. Discrimination: we must be able to tell the difference
if 0, white if 255 and grey if in between.
between wavelengths. o Hue: chromatic aspect of light (each point is a
Problem of univariance: the output of a single
photoreceptor is ambiguous (several combinations o Saturation: chromatic strength of a hue (white
of wavelength intensities will produce the same is zero saturation, pink is more saturated, and
respnonse in the photoreceptor). red is full saturation)
In the dark, because only one type of rod cones o Brightness: perceptual consequence of physical
exist (rhodopsin), we can’t discriminate colours.
intensity of light.
This shows that color is psychophysical, not Non-spectral hues are those mixtures that can
physical. Problem of univariance persists. never be represented with a single wavelength
Trichromatic solution: trichromatic theory of color (e.g. purple magenta between red and blue).
vision (trichomacy): color of any light is defined by Opponent color theory: four basic colors in two
relationships of the outputs of 3 cones (aka young-
opponent pairs: red vs. green and blue vs. yellow.
Helmholtz theory). This can be proved with the hue cancellation
The relation among the cones (twice M as S) does theory (we can cancel the yellowness in a color by
not change across light intensities, and that’s why adding the right amount of its opposite color). This
we see the same color in different light settings. also shows unique hues (any of the 4 colors –
Metamers: mixture of different wavelengths that yellow, red, green and blue – that can be described
looks identical. Color mixture is mental (it’s not the with only a single color term).
sum or the average), not the change in physics of
(L+M)-S: results in cells that respond maximally
light. along an axis extending from purple to yellow-
History of trichromacy: It started with Newton green hues. L-M: L is red and M is blue-green
suggesting that a prism breaks sunlight into the (those are the cardinal directions)
spectrum of hues. Matthews also found that only Cells in visual cortex: LGN has cone-opponent cells
three colours of light are needed to match any
so the transformations that produce color-
reference color. opponent processes that support appearance are
Additive color mixture (adding wavelengths likely to be found in visual cortex.
together) vs. Subtractive (if A and B mix, some light o Double-Opponent Cell: found in visual cortex,
will be subtracted by each, and what we see is the R+/G- in center and R-/G+ in surround or vice
remainder). An example of subtractive: color filters versa; conveys information about chromatic
From retina to brain: nervous system looks at edges
differences of activities in the 3 cone types. It o Single-Opponent Cell: R in center and G in
computes 2 differences: surround; conveys information about the color
o L-M: suited to notice differences between
different amounts of blood on skin. We can’t separate perceptual processes in cortical combination (blue-green), and has to be common.
anatomy, so color is everywhere, not just in a In English, we have 11 languages. Cultural
specific region. relativism (each group was free to create its own
linguistic map of color space) was created as a
Best evidence to say that color does have a
specialized brain area is achromatopsia (loss of result but Berlin and Kay found that most of the
color vision after brain damage). colors in other cultures resemble our organization.
Adaptation can be color specific as can be seen The question now is: do cultures who have 2 colors
through the phenomenon of negative afterimages perceive colors differently? The Dani tribe has only
2 colors: mola (light-warm) and mili (dark-cool).
(visual image seen after the physical image was
removed). The first image is called the adapting They were shown a coin then asked which of 2
stimulus, and what comes after is the negative pictures resembled the coin color more. Their
afterimage. boundaries were just like ours. Another group that
Do We See Colour the Same Way? was studied was the Bernimo who had nol/wor
(distinction lies in the middle of colors we
o Yes-version: Performance on standard measures of
color vision will be the same; 2 lights are categorize as green). When they did the color test,
metametrically matched. You may have slight they did better on the nol/wor boundary than on
differences due to factors like age (which makes the blue/green boundary (opposite effect for
lens yellower). English-speakers). Similar results were found with
Russia’s dark-goluboy and light-siliy. So now we
o No-version: Color-blindness occurs in 8% of males
and 0.5% of females because M- and L- cones are don’t know if culture can affect perception.
found on the X-chromosome. There are different From the Color of Lights to a World of Color
types of color-blindness because of (1) the cone Color contrast effects (color of one region induces the
that may be affected and (2) type of defect (either opponent color in a neighboring region) vs. Color
photopigment is color-anomolous – all types Assimilation effects (2 colors bleed into each other,
present, but 2 will be similar – the cone is just taking on some chromatic quality of the other).
missing, or problems in cortex): Unrelated colors (colors that can exist in isolation) vs.
Deuteranope: someone without M-cones (2D) related colors (colors that can be seen only in the
Protanope: someone with no L-cones (2D) context of others; can’t be isolated; e.g. grey will be
Tritranope: someone with no S-cones (2D) seen is white in darkness).
Cone Monochromat: only one type of cone in Color constancy: tendency for colors to appear
the retina (1D; only in gray) relatively unchanged in spite of changes in lighting. This
Rod Monochromat: missing cones altogether is difficult for the visual system to perform because the
(worst type; poor acuity and difficulties seeing illuminant (light that eliminates a surface) is not
in normal daylight; only have rods) constant. The spectral reflectance function for a surface
Achromatopsia: world drained of color is the percent of each wavelength that is reflected from
a specific surface. The spectral power distribution of
Agnosia: see something but not know what it is
Anomia: inability to name things (like color) the surface is the relative amount of light at different
We know of this because of studies on people that visible wavelengths. The light that reflects to our eyes is
are color-blind in only one eye and who can compare product of the surface and illumination. The problem is
what they see between both. that even though the lights are different, and produce
different lights, we still see the original surface similarly,
Maybe-Version: With exception of color-deficient
individuals, we agree about the basic color, but not such that we can recover the color of the surface as well
specificity (is it orangish-yellow or yellowish- as know something about the illuminant. Our
orange). The number of `basic’ color terms differs perception is actually based on assumptions that limit
across culture. The ‘basic’ colors have to be not a the possible answers to our question: how much of the
surface area is there really. Together, the below
substance name (bronze/lavender) or a assumptions our system makes help us see better The trick to the different cones we have is that they can
(Bayesian theory). act like each other with the right oil filter placed in front
Bright-is-White theory: in a complex scene, we would of it. Birds and chicks have this too. This can also be
assume that the brightest color is white, and after that seen in fireflies. They are bioluminescent (make their
is grey. This is wrong because what if we were in a own light) so the light, with a specific filter will make
dark room with only red and blue spots? We still see one firefly’s light brighter, so that it’s easily recognizable
red and blue, but according to this we’d see white. by its species.
Assumptions can be made about illuminants. This Animals, like us, are interested in surface’s properties
(not the light itself), so color constancy is important for
won’t always work because highly unnatural light
sources (like those in club) make the world seem them too.
unnatural This information about animals reminds us that color is
Assumptions about the surface: real surfaces tend to a mental concept, not a physical one.
have reflectance (% of light hitting a surface and
reflected or absorbed by it). The whitest surface rarely
reflects more than 95%, yet we still see it as white. The ability to perceive and interact with the structure of
Assumptions about the structure of the world: sharp space is a fundamental goal of our visual system.
borders are almost always the boundaries of 2 objects, Realism: philosophical view arguing that there is a real
not 2 light sources. This helps us with shadows world to sense.
Positivism: a view that all we really have to go on is the
(because our system knows that shadows are a result
of light intensity changes, not hue changes). evidence of the sense so the world may be nothing but
an elaborate hallucination.
An experiment to show that vision is the nervous
system’s best guess about what’s happening in the The geometry of our real world is Euclidean (parallel
world: researchers showed participants a card (half lines remain parallel, objects maintain the same size and
white and half red). Then they folded it so that the shape as they move in space, and internal angles of a
colours faced each other, and the light from the red triangle are 180…). Yet, the geometry of our retinal
images is non-Euclidean. It becomes so when the world
reflected unto the white, making the white look like
pink. Because the visual system knows about reflection, is projected onto our curved 2D surface of the retina.
it still saw it as white. When this was done again, but So, if we want to experience the Euclidean world, we
with a the card looking like a roof instead, the visual have to do so from a non-Euclidean point view (through
system didn’t have an explanation as to why there was our 2 retinas).
Our retinas project slightly different images because of
pink if this was ‘supposed’ to be a roof, so observers
saw it as pink. their different location. Why do we have 2? Same
Color Vision in Animals reason we have 2 lungs, kidneys… so that we can still
Color vision has evolved several times, which means it function if we lost one.
must be important to cost so much. Across the animal The location of our eyes causes us to see a shared
image of 110 degrees, and an additional 40 degrees for
kingdom, there seems to be 2 important reasons for it:
Sex and eating. Flowers put pretty UV patterns (we each side (190 in total). We can also see 60 degrees
can’t see; they’re beyond our range because they’re too from center to up, and 80 degrees from center to down.
short) too attract bees to give them food in place of Rabbits have a 360 visual degree.
pollination. Some flowers have really long wavelength Binocular (2 eyes) visual fields give predators an
advantage to spot small, fast-moving objects. Binocular
patterns (beyond our range too) to attract bats, but not
through photoreceptors, through thermoreceptors in summation (combined signals from each eye in ways
the snout. Animals have pretty colorful patterns too to that make performance on many tasks better than with
attract mates. one) may have served an evolutionary purpose).
We are trichromats because we have 3 cones. Dogs are Binocular disparity: difference between 2 retinal
images of the same world. This is the basis for
di and chicken are tetrachromats. stereopsis (use disparity as a cue to depth but not the same image from anywhere we sit. This is because
completely necessary for it), which is not possible with our visual system takes the orientation of the flat
monocular vision. surface into account, allowing them to understand
Depth cues: information about the 3 dimension of that this is just a picture, not reality. When there’s
visual space (monocular or binocular) enough context, our visual system can also account for
Monocular Cues to 3D Space distortions in pictures.
Monocular depth cues are those that are available even Anamorphosis: use of rules of linear perspectives
when viewed with one eye alone. (pushed to an extreme) to create a 2D image so
Occlusion: a depth cue, that gives information about distorted that it looks correct only when viewed from
relative position of objects. It’s the most reliable, a special angle or mirror that encounters the
except in cases with exceptional viewpoints. Occlusion distortion. This shows that our ability to cope with
is a nonmetrical depth cue (just gives us the relative distortion is limited.
orderings of occluders and occludees) as opposed to Motion parallax: based on head movement.
metrical depth (provides information about distance Geometric information obtained from an eye in 2
in 3D) different positions at 2 different times is similar to
Relative size cue: A comparison of size between items information from 2 eyes in different positions in the
without knowing absolute size of either one. This is head at the same time. Objects closer to you will
because our visual system knows facts about change position more than farther objects as you
projective geometry (descriptions of transformations move down a track. If you sit under a tree with one
that occur when 3D world is projected into a 2D eye open and look up, the branches will all look that
surface. they’re at the same distance. If you open both eyes,
Texture gradient: items of the same size form smaller stereopsis (binocular depth cue) will help you with the
images when farther away. depth. However, with one eye, depth can also be seen
Relative height: observation that objects at different if you moved your head (with one eye open) from side
distances from viewer on ground plane will form to side. This provides relative metrical depth cues.
Motion parallax only works when heads move, not
images of different heights.
Familiar size: based on knowledge of the typical size when eyes move.
of objects like humans or pennies. Accommodation: process by which eye changes focus,
Relative metrical depth cue (could specify that an in which lens gets fatter as directed toward nearer
object is twice as far as another without providing objects.
absolute distance, for example; relative size and height) Convergence (ability of the 2 eyes to turn inward to
vs. absolute metrical depth cue (provides quantifiable place 2 images of a feature on corresponding locations
information about distance in 3D; familiar size cue). in the 2 retinal images. It reduces the disparity of that
Aerial Perspective (haze): objects far away are subject feature to 0) vs. Divergence (ability of 2 eyes to turn
to scatter and appear fainter and less distinct. This is outward to place 2 images of a feature in the world on
because of the understanding that light is scattered by corresponding locations in 2 retinal images. This also
the atmosphere. reduces disparity to 0). These (with accommodation
and familiar size) are the only ones that tell us
Linear Perspective: based on the rules that determine
lines in the 3D space projected onto a 2D image. absolute metrical depth
Vanishing point: apparent point at which parallel lines information.
receding in depth converge. Binocular Vision and Stereopsis
Pictorial depth cues: all the aforementioned depth Corresponding retinal points:
geometric concept stating that 2
cues! It’s used by artists to depict 3D on their 2D
paintings/pictures. In theory this means that there’s points on retina of each eye where
only one position available to view the picture monocular retinal images of a single
correctly, but we can sit in the theatres and still see object are formed at the same distance from the fovea in each eye. Vieth-Muller circle: location of objects printing (images are digitally split and interleaved for
whose images fall on geometrically corresponding right, left, right, left… at a fixed spacing.
points in 2 retinas. Horopter: location of objects whose We don’t get adequate disparity from objects at large
images lie on corresponding points (0 disparity). distances because our eyes are only a few inches apart.
Diplopia: double vision; if visible in both eyes, stimuli Sterestopic displays are now used in radiology for
falling outside of Panum’s fusional area (region of space cancer detection, as well as in the military for a better
in front of and behind the horopter within which understanding of the ground below (seen from an
binocular single vision is possible) will appear diplopic. airplane) and its depth.
What the visual system assumes is that the greater the Correspondence problem: problem of figuring out
disparity between retinal images on both eyes, the which bit of image in left eye should be matched with
greater the distance in depth of the object from which bit in the right eye. The system uses heuristics for
horopter. achieving correspondence:
Crossed (signs of disparity created by objects in front of o Blur image: you won’t be able to see the actual
the plane of fixation (horopter); object is on left in right image, but you’ll see fewer items and can find out
eye and right on left eye) vs. Uncrossed (signs of which blob corresponds to which one in the other
disparity created by objects behind the plane of eye
fixation; object is on left in left eye and right on right o Uniqueness: a feature in the world is represented
eye) disparities. exactly once in each retinal image.
Stereoscope: device that presented one image to one o Continuity constraint: except at the edges of
eye and a different one to the other. This instrument objects, neighboring points in the world lie at
proved that the visual system treats binocular disparity similar distances from the viewer.
as a depth cue, regardless of whether disparity is Convergence doesn’t happen until the striate complex,
produced by actual or stimulated images. Stereoscopes where most neurons are binocular (has 2 receptive
are helpful, but the same experience can be seen if you fields, one in each eye). These receptive fields are
train yourself in free fusion (technique of converging – generally similar and share the same spatial-frequency
crossing- or diverging the eyes in order to view a and orientation tuning. So, they’re suited for matching.
stereogram without a stereoscope). Some neurons respond best when object lies within
Stereoblindness: 3-5% of population have a lack of horopter and others when the images occupy different
stereoscopic depth perception. This is usually a positions on the retinas (these neurons are tuned to
secondary effect of childhood visual disorders like binocular disparity).
strabismus (2 eyes are misaligned). Stereopsis can be metrical or not: nonmetrical might tell
Julesz believed that stereopsis might be used to you that a feature lies in front of or behind the plane of
discover objects and surfaces. He made random dot fixation (these neurons were found in V2 and higher).
stereograms (made of a large number of dots, that are These neurons respond to either near (crossed) or far
randomly placed. In monocular views, they show (uncrossed) stimuli. Metrical stereopsis (hyperacuity)
nothing, but stereoscopically, they are Cyclopean has thresholds smaller than size of a cone. Dorsal
stimuli – stimuli that are defined by binocular disparity (where) pathway has metrical stereopsis and ventral
alone). (what) pathway has nonmetrical stereopsis.
For 3D movies, the film is shot from 2 videos, 2 inches Combining Depth Cues
apart. Then the viewers wear glasses that can be Helmholtz first noticed that combining visual cues is
anablyphic (present different color filters to the eyes) or what makes our visual system works. He called it
more recently, interference (filters of different ‘unconscious interference’ but it is now known as the
wavelengths). Also, LCD (special crystal liquids) can be mathematical method: Bayesian Approach (a way of
used to enable the presentation of images to each eye formalizing the idea that our perception is a
be synchronized with images in the movie/TV. All these combination of current stimuli and our knowledge
need special glasses, but a new method is lenticular about the conditions of the world – P(A|O) – P(A) x P(O|A)/P(O) – probability (p) that the world is in a adult monkeys, so the infants’ blindness must be
particular state (A) given a particular situation). because of something at V2 or beyond.
Example: you see 2 pennies, the smaller one lies behind Critical period: period during early visual development
the other. There are infinite possibilities of what’s when normal binocular visual stimulation is required for
actually there: are they equidistant but one is smaller normal cortical development. This was found to be at 3-
and has a piece bitten off, is the smaller one farther, is 4 months in cats and monkeys, but in humans, we can’t
the bigger one closer… Our system uses prior study them like that, so we look at existing disorders:
knowledge to determine the most likely possibility. Stabismus: misalignment of the eyes so that a
Ideal observer analysis: theoretical observer with stimulus is found on the fovea of one, but not on the
complete access to best available information and other eye (3% incidence rate).
ability to combine the different sources. o Esotropia: one