Transcript Visual system 3

Color vision
3 forms of Rhodopsin are sensitive to different parts of the spectrum:
Sensation of whole spectrum of colors provided
by exciting differently the 3 cone types
Trichromatic theory of color vision
• Young and Helmholtz proposed that humans have
3 kinds of photoreceptors that work together based
on observation that any color of light can be attained
by mixing various amounts of 3 colors of light.
Opponent Process Theory of Color Vision
• Proposed by Hering
• Based on:
- Some colors don’t blend (e.g. no reddish green)
- Observation of negative afterimages
* Trichromatic theory can’t explain these phenomena
2 kinds of color sensitivity in ganglion cells
red
opposes
green
blue
opposes
yellow
3 types of receptive fields with complementary colors
Blue on,
yellow off
Red on,
green off
green on,
red off
Opponent Process Theory of Color Vision
RETINAL COLOR CODING
cones
440
530
Red light “stimulates”
560
red cone
Red cone “stimulates”
red/green ganglion cell
ganglion cells
signals red
RETINAL COLOR CODING
cones
440
530
blue light “stimulates”
560
blue cone
blue cone “inhibits”
yellow/blue ganglion cell
ganglion cells
signals blue
RETINAL COLOR CODING
cones
440
530
yellow light “stimulates”
red and green cones
560
equally
Red and green inputs to
red/green cell cancel
red and green sum to
“inhibit” blue/yellow cells
ganglion cells
signals yellow
RETINAL COLOR CODING
cones
440
ganglion cells
530
560
Accordingly, we can see:
orange
reddish-yellow
purple
reddish-blue
turquoise
greenish-blue
and
lime
greenish-yellow
but
we cannot see
reddish-green
or
bluish-yellow
Opponent Process Theory of Color Vision
Ganglion Cell responses
Color Assimilation
(The Von Bezold spreading effect)
The blue or yellow (depending on the diagonal)
areas adjacent to the red squares influence the
appearance of the red squares. The yellow make
the red appear lighter and the blue make the red
appear darker.
This phenomenon is the opposite of a contrast
effect where one expects near by colors to
accentuate the differences between adjacent
areas.
Motion Perception
• Motion perception is the process of inferring the speed and direction of
elements in a scene based on visual ,vestibular and proprioceptive (expected
movement of objects on the retina when we move) inputs.
• Although this process appears straightforward to most observers, it has proven to
be a difficult problem from a computational perspective, and extraordinarily difficult
to explain in terms of neural processing.
• Mechanism: delay
http://www.physpharm.fmd.uwo.ca/undergrad/sensesweb/L4Motion/L4Motion.swf
Motion Perception
• First-order motion perception: the perception of the motion of an object that
differs in luminance from its background.
• Second-order motion perception: the moving contour is defined by contrast,
texture, flicker or some other quality that does not result in an increase in
luminance or motion energy. Second-order mechanisms have poorer temporal
resolution and a weaker motion aftereffect (MAE).
• There is much evidence to suggest that early processing of first- and secondorder motion is carried out by separate pathways. First and second-order signals
appear to be fully combined at the level of Area MT of the visual system.
V5 / MT
• The most prominent region of motion perception – the MT – was found in owl
monkeys at the posterior end of the Middle Temporal gyrus. In humans the
equivalent area is the ITS (Inferior Temporal Sulcus).
• With no MT the sense of motion is lost and we perceive a series of stills.
• MT is tuned to direction and speed of movement. It is divided into direction
columns.
• Does not identify objects due to poor resolution + has no color vision. Need to
combine with “WHAT” stream.
• Inputs to MT: Magnocellular LGN, V1, V2, dorsal V3
• MT sends its major outputs to areas located in the cortex immediately
surrounding it, including areas FST, MST and V4t (middle temporal crescent).
• Activity is dependent on attention.
MST
MST = Medial Superior Temporal Area
The Lateral part – MSTl:
• Senses movement.
• Small receptive fields.
The dorsal part (MSTd):
• Senses visual motion produced when we move. During movement the
background produces an optic flow pattern on the retina, which activates
these cells.
• MSTd is organized in columns tuned to different patterns of optic flow.
• Large receptive fields.
Corollary Discharge
An object moves on the retina when
either the object or the eye moves.
To distinguish between the two options
we use information on eye movement =
corollary discharge
Corollary discharge is characterized as an
afference copy (first suggested by von
Helmholtz) of an action command used
to inhibit any response to the self
generated sensory signal
The Aperture Problem
• When the ends of lines are not apparent, we cannot conclude the
direction of movement.
• Motion sensitive cells are faced with the same problem because of
their receptive fields.
• The system has to imagine where the lines end to guess the direction
of motion.
• Area MT differs from V1 in its response to 2 sets of lines
Feature Integration Theory (Anne Treisman)
• Objects are processed pre-attentively at a feature level first (simple features
are registered in parallel in specialized subsystems)
• Focused attention is needed to serially scan, integrate, and bind these
features into objects.
• Different kinds of attention are responsible for binding different features into
consciously experienced wholes.
• Treisman concluded from many experiments that color, orientation, and
intensity are primitive features, for which feature search can be performed.
• Two kinds of visual search tasks:
1. Feature search - performed fast and pre-attentively for targets defined by
primitive features.
2. Conjunction search – a serial search for targets defined by a
conjunction of primitive features. Much slower and requires conscious
attention.
Feature Integration Theory (Anne Treisman)
Evidence for this theory:
Blue
Green
Red
1. Illusory conjunctions: when attention is overloaded, illusory conjunctions are formed.
Veridical binding of features requires focused attention.
2. Initial vision identifies elements without registering their precise location
3. Popout of primitives in visual search: when the target element is defined by a single
distinctive feature (a large difference in color, orientation, or size), its detection is rapid
and parallel (independent of the number of distractors).
A,B - pop-out happens if the odd
element differs only in a single feature
Search with serial focused attention is
required if:
C - the odd element differs only in a
conjunction of features
E,F - the odd element differs only in the
spatial arrangement of element parts
D - the difference is small
Feature Integration Theory (Anne Treisman):
Problems with the theory
Associating initial feature search with low cortical areas has been
questioned in a number of ways:
1. Low-level neuronal mechanisms have precise position information, while
pop-out does not
2. Low-level neuronal mechanisms discriminate fine orientation or color
differences that do not pop-out.
3. The presumed “automaticity” of pop-out: recent studies indicate that
attention, albeit spread, is required for feature search. It depends on
limited resources, since simultaneous performance of another task may
interfere with it
4. While some of the simple features that pop-out (orientation, color,
motion, and depth) are represented in V1, others are not, including
closure and geometric shape (circles versus rectangles).
Reverse Hierarchy Theory
(Merav Ahissar and Shaul Hochstein)
• RHT assigns the pop-out phenomenon to initial perception at high-level areas
using their large receptive fields. Later feedback reentry to low levels slowly
adds details available in the small specific receptive fields found there.
• Vision at a glance reflects high-level mechanisms, while vision with scrutiny
reflects a return to low-level representations.
• Based on the apparent disparity between our extremely rapid capture of the
conceptual gist of a scene together with our blindness to its details.
Since the whole is surely built of its parts, how is it that the parts remain
unknown, while the whole becomes accessible?
• Feature search reflects high cortical level activity based on large, spreadattention receptive fields. This leads to position- and size-invariant feature
detection, as suggested by the Feature Integration Theory.
Reverse Hierarchy Theory:
Pop Out
• Pop-out is another manifestation of rapid but high-level vision at a
glance. Thus the features that pop out and their characteristics will match
those of high- but not low-level receptive fields.
• Many high-level features pop out:
- 3D features (B) and depth from shading (D)
- facial expressions
- parts and wholes
- perceptual groups
Reverse Hierarchy Theory
• Both conjunction search and fine detail discrimination search depend on
low-level area neurons.
• When two distractors are present, search is hard when the target lies
between them in feature space but is easy otherwise:
- orange doesn’t pop-out of red
and yellow
- 45° pops out of 15° and 315°
distractors but not of 15° and
75° distractors
Reverse Hierarchy Theory
• Features that pop-out are better described in terms
of categories than as measurable spatial
characteristics.
• For a target object to pop-out it must be represented
by a neuronal population that does not overlap with
the population representing the distractors.
Since face representation is largely separated from
that of other objects, they are expected to pop-out.
• Circles pop-out of squares because faces pop-out of
houses, not vice versa.