Transcript Lecture 4: Visual Perception October 1, 1999

Visual Computation
I. Physiological Foundations
II. Elements of Visual Computation
III. Synchronized Oscillation in Visual Cortex
I. Physiological Foundations
Visual perception:
“What you see is not what is out there in the
outside world, but instead what has been
created all inside of your head as the nervous
system’s response to external stimuli.”
“Visible” Spectrum
Not all ranges of the light spectrum is
perceived. Why?
The Eye
The Retina
Steps of Visual Computation
S1: Rod & Con cells (light detection)
S2: Horizontal & Bipolar cells (preprocessing)
S3: Ganglion cells (preprocessing)
S4: Lateral geniculate nucleus (LGN) cells
(relay station)
S5: Visual cortex (conscious perception)
Rods and Cons
Rods:
- Discriminate B/W brightness in low illumination
- Contribute to peripheral vision
Cons:
- Discriminate Colors
- Contribute to central vision
Three Types of Con Cells
Each wavelength gives rise to a different
combined activity pattern of the three cone
types, thus giving experience of different
colors.
Why three cone types?
Why not one or two or even several?
What would be like one-cone perception?
Color-misnomer phenomenon.
One Type of Con Cells
Two Types of Con Cells
II. Elements of Visual Computation
1. Receptive field (RF)
2. ON-center/OFF-surround RF
3. Edge detector
4. Orientation detector
5. Location-invariant detector
6. Ice-cube model of visual cortex
7. Color computation
1. Receptive Field
The receptive field (RF)
of a neuron is the area
of retina cells that
trigger activity of that
neuron.
2. On-center/OFF-surround RF
3. Edge Detector
4. Orientation Detector
(Huble & Wiesel, 1959)
Hypothetical Wiring Diagram
5. Location-invariant Detector
6. Ice-cube Model of Visual Cortex
Orientation Selectivity of Cortical Neuron
Face-detecting “Grand Mother” Neuron
7. Color Computation
Transformation: (B,G,R) -> (W, B-Y, R-G) -- no information loss, then why?
Take-home Message:
“What you see is not what is out there in
the outside world (ie., not like “taking a
picture”), but instead a result of visual
computation -- only those computations that are critical
for survival, shaped by the evolution.”
What about Illusions?
-> “Unintended” consequences of the computation
III. Synchronized Oscillation in Visual Cortex
1. Sensory Binding Problem
2. Neural Assemblies
3. Synchronized Firing
1. Sensory Binding Problem
In the brain, spatially different areas are functionally
specialized for processing visual attributes such as
shape, color, orientation, and direction of motion.
Binding Problem:
How does the brain then “bind” together the sensory
attributes of an object to construct a unified perception of
the object?
Binding Problem
2. Neural Assemblies
Neural assembly: a group of neurons whose activities
are highly correlated/inter-dependent to one another.
3. Synchronized Firing
Neural synchrony: in which neurons in a neural assembly
tend to fire in phase-lock fashion or in unison.
Global Synchronization
via long-range excitatory connections
… Meaning that the following was the input stimulus,
Synchronized Neural Assembly Coding
->Solution to the Binding Problem
Two or more assemblies, each of which is tuned to a
particular feature/attribute present in a stimulus object,
contribute to the representation of that object by jointly
firing synchroniously and simultaneoulsy.
Biological plausibility:
well-documented (e.g., Gray et al, 1989).
Vision is a computation.