1 computational vision csci 363, fall 2012 lecture 3 neurons central visual pathways see reading...

18
1 Computational Vision CSCI 363, Fall 2012 Lecture 3 Neurons Central Visual Pathways See Reading Assignment on "Assignments page"

Upload: valerie-moody

Post on 12-Jan-2016

213 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: 1 Computational Vision CSCI 363, Fall 2012 Lecture 3 Neurons Central Visual Pathways See Reading Assignment on "Assignments page"

1

Computational Vision

CSCI 363, Fall 2012Lecture 3

NeuronsCentral Visual Pathways

See Reading Assignment on "Assignments page"

Page 2: 1 Computational Vision CSCI 363, Fall 2012 Lecture 3 Neurons Central Visual Pathways See Reading Assignment on "Assignments page"

The NeuronMajor parts of a neuron:

1) Cell body2) Axon3) Terminal4) Synaptic cleft5) Dendrites

Page 3: 1 Computational Vision CSCI 363, Fall 2012 Lecture 3 Neurons Central Visual Pathways See Reading Assignment on "Assignments page"

3

Action Potentials

Action potential

Na+ current

K+ current

1. The resting potential is an electrical potential (from -40 to -110 mV) across the cell membrane.

2. Excitatory input causes the potential to decrease (depolarize).

3. This causes a voltage-sensitive mechanism to set off a rapid change in electrical potential in the membrane.

4. Action potentials move rapidly down the axon to the axon terminal.5. Action potentials require energy.

Page 4: 1 Computational Vision CSCI 363, Fall 2012 Lecture 3 Neurons Central Visual Pathways See Reading Assignment on "Assignments page"

4

Neural Signaling1. Input to the cell causes depolarization of the cell body to threshold.2. An action potential propagates down the axon to the terminal.3. Transmitter is released, diffuses across the synaptic cleft to the postsynaptic

cell and binds to receptors on the postsynaptic cell.4. The transmitter causes an electrical (positive or negative) response in the

dendrite of the postsynaptic cell.5. The electrical signal in the dendrite spreads passively to the cell body.6. Electrical signals from many dendrites combine at the cell body. If excitation

outweighs inhibition, the cell may reach threshold causing another action potential.

7. The cycle begins again in the next cell.

Page 5: 1 Computational Vision CSCI 363, Fall 2012 Lecture 3 Neurons Central Visual Pathways See Reading Assignment on "Assignments page"

5

Complexity of Neural Processing

•One neuron may connect to 1000 post-synaptic neurons.

•One neuron may receive 10,000 inputs from other cells.

•Lots of computation takes place locally within the dendrites.

•There are approximately 1011 neurons in the brain.

Things to bear in mind:

Page 6: 1 Computational Vision CSCI 363, Fall 2012 Lecture 3 Neurons Central Visual Pathways See Reading Assignment on "Assignments page"

6

Simplified Model Neurons

Typical neural network models use very simplified model neurons:

1) The response is modeled as the rate of action potential firing. This is a graded response with a threshold and saturation.

2) The neurons are generally homogeneous. There are no differences in shape or transmitter effects.

3) A single neuron can have both excitatory and inhibitory effects on the postsynaptic cell.

4) There is no systemic modulation of neurons (e.g. from hormones).

Page 7: 1 Computational Vision CSCI 363, Fall 2012 Lecture 3 Neurons Central Visual Pathways See Reading Assignment on "Assignments page"

7

Central Visual Pathway Overview

Processing of visual information in primates is accomplished in stages along the visual pathway:

Retina

Lateral Geniculate Nucleus

Striate Cortex

Extrastriate cortex

Page 8: 1 Computational Vision CSCI 363, Fall 2012 Lecture 3 Neurons Central Visual Pathways See Reading Assignment on "Assignments page"

8

The Eye

Page 9: 1 Computational Vision CSCI 363, Fall 2012 Lecture 3 Neurons Central Visual Pathways See Reading Assignment on "Assignments page"

9

Processing in the Retina

Light is absorbed by the photoreceptors in the retina.

Some processing is performed in the retina itself.

The retinal ganglion cells are the final stage in the retinal processing. They send axons out of the eye to the LGN.

Retinal ganglion cells have receptive fields with center-surround spatial structure.

+-

On center/off

surround

-+

Off center/on

surround

Page 10: 1 Computational Vision CSCI 363, Fall 2012 Lecture 3 Neurons Central Visual Pathways See Reading Assignment on "Assignments page"

10

Projections from the RetinaAxons from the retina project to three areas:

1) Pretectal area: a midbrain area that controls pupillary reflexes.

2) Superior colliculus: controls saccadic (rapid) eye movements.

3) Lateral geniculate nucleus (LGN) of the thalamus: This is the principal pathway, which projects to visual cortex.

4) Some of the axons from the retinal ganglion cells cross over to the opposite side of the brain before reaching the LGN (at the optic chiasm).

Page 11: 1 Computational Vision CSCI 363, Fall 2012 Lecture 3 Neurons Central Visual Pathways See Reading Assignment on "Assignments page"

11

The Optic Chiasm

Page 12: 1 Computational Vision CSCI 363, Fall 2012 Lecture 3 Neurons Central Visual Pathways See Reading Assignment on "Assignments page"

12

The Lateral Geniculate Nucleus

The LGN is layered.

Input from the two eyes largely separate.

Projection from retina is retinotopic.

Cells respond like RGCs.

Page 13: 1 Computational Vision CSCI 363, Fall 2012 Lecture 3 Neurons Central Visual Pathways See Reading Assignment on "Assignments page"

13

Primary Visual CortexPrimary Visual Cortex (also known as Striate Cortex or V1) is the

first cortical area in the visual pathway.Hubel and Wiesel (1950's and 60's) were the first to describe

properties of V1 cells. (Movie)

They described 3 types:Simple cells: Elongated Receptive fields. Orientation selective.

Defined regions of excitation and inhibition.

Complex cells: Also orientation selective. No well defined regions of excitation and inhibition.

Hypercomplex cells: End-stopped.

Page 14: 1 Computational Vision CSCI 363, Fall 2012 Lecture 3 Neurons Central Visual Pathways See Reading Assignment on "Assignments page"

14

Ocular Dominance Columns

•The cortex is a folded sheet of cells, about 2 mm thick.•The cells form layers (6 layers in primary visual cortex).•If move perpendicular to the surface of the cortex, cells will respond primarily to input from one eye (ocular dominance).•The pattern of responses forms columns of ocular dominance.

Page 15: 1 Computational Vision CSCI 363, Fall 2012 Lecture 3 Neurons Central Visual Pathways See Reading Assignment on "Assignments page"

15

Orientation Columns

Preference for a given orientation also has a columnar structure:

Page 16: 1 Computational Vision CSCI 363, Fall 2012 Lecture 3 Neurons Central Visual Pathways See Reading Assignment on "Assignments page"

16

Retinotopic organization

The visual field maps onto visual cortex in a systematic way.

More of V1 is devoted to processing the central visual field than to the periphery.

Page 17: 1 Computational Vision CSCI 363, Fall 2012 Lecture 3 Neurons Central Visual Pathways See Reading Assignment on "Assignments page"

17

Extrastriate Visual Areas

This map shows a flattened cortex with the known visual areas mapped onto it.

•There are a large number of distinct visual areas (probably at least 20).•Each area appears to have a specific function.•The areas show a roughly hierarchical organization (although most areas have reciprocal connections).

Page 18: 1 Computational Vision CSCI 363, Fall 2012 Lecture 3 Neurons Central Visual Pathways See Reading Assignment on "Assignments page"

18

Two Major Processing Streams

There appear to be 2 major processing streams (although there are cross connections between them):

1. The Dorsal Stream:Includes areas MT, MST, VIP, 7a, etc.Processes motion, stereo, spatial relationshipsThe "where" pathway.

2. The Ventral Stream:Includes areas V4, IT, etc.Processes color, form, objects.The "what" pathway.