Spatial Vision

1
Spatial Vision
3. Impact of neural processing on spatial vision
2
Chapter 5 Spatial Vision Part C Role of
lateral inhibition
3
Role of receptive field (RF) excitation and
inhibition.
Center-surround Receptive Field with Lateral
Inhibition
center excitation
EI
excitation
inhibition
Receptive Field schematic
surround inhibition
4
Response Excitation Inhibition Therefore,
light falling in the RF inhibitory surround will
reduce the neural excitatory response to light
falling in the RF center.
Effect of inhibition
Maximum response will therefore occur when there
is a lot of excitation, but no inhibition.
What type of stimuli will produce (1) lots of
excitation, but no inhibition, and (2) high
levels of both excitation and inhibition? What
will be the perceptual consequences of inhibition?
5
Examine Four Phenomena that cannot be explained
by sampling and summation and optics.
1. Band-pass nature of CSF 2. Simulataneous
Contrast Effects 3. Hermans Grid
illusion 4. Mach Bands
6
CSF
?
Peak CS
?
?
Contrast Sensitivity (1/contrast threshold)
High SF cut-off
SF _at_ peak
low
high
medium
Spatial Frequency (c/deg)
7
Simultaneous Brightness Contrast
Notice that the equally luminous small squares
and circles do not appear equally bright.
8
In this example we see that the requirement for
single surfaces to have the same brightness can
over-ride the local contrast effect.
With the red line in place, the two half annuli
have different brightness because of their
surrounding light levels. But, when the red line
is removes, and the annulus becomes a single
object, we see it as having only one brightness
level.
9
Hermans Grid
Can you see dark patches at the intersection of
the white lines? Are they missing at the
fixation point?
10
CSF
11
Simultaneous Brightness Contrast
Compare the response of the RF centered on the
two patches. Which one has the most inhibition
and thus lowest response?
12
Explanation for dark patches, and their absence
in the fovea where RFs are smaller.
Hermans Grid
13
Explanation square B is in a shadow where all
surfaces have lower luminances, and in this
example, the darks squares outside the shadow
have the same luminance as the light squares
within the shadow, but because of brightness
constancy, we always see the bright squares as
bright. Said another way, irrespective of the
overall light level (in or out of shadow),
brightness is determined by the local contrasts.
14
Another example showing the simultaneous
brightness illusion
15
Heterogeneity of retinal RF properties a b,
X Y, Parvo Magno, parallel pathways
Magno projection
Alpha cells large dentritic fields, Y cells,
high sensitivity, low resolution, highly
transient, short latency, spectrally non-opponent
a
Beta cells small denritic fields, X cells, lower
sensitivity, more sustained, longer latency,
spectrally opponent.
CS
b
Parvo projection
Spatial Frequency
16
Change in RF structure from retina to cortex
Retina concentric (circular), tri-lobed
excitation
Ecitatory Center
large
Inhibitory Surround
small
inhibition
17
Cortex Orientation-selective, multi-lobed
Multiple-lobed RFs respond to a select range of
SF (SF tuned)
Small High SF
Large Low SF
excitation
inhibition
18
Cortical coding of spatial scale
Band-width of individual cortical neurons
CS
Spatial Frequency
19
CS varies slightly with orientation.
CS
Orientation
Reduced Sensitivity to oblique contours the
oblique effect
20
2. Orientation and SF selectivity
Blakemore Campbell Contrast Adaptation
Experiment
1st measure CS
2nd prolonged exposure to high contrast gating
(keep eyes moving)
3rd measure CS
CS
Adaptation frequency
Result sensitivity is reduced only to those SF
near to the adapting SF (a spatial
frequency-selective effect)
Spatial Frequency
21
Same contrast adaptation experiment, but examine
the spread of spatial adaptation across
orinetation.
Result sensitivity is reduced only to those
orinetations near to (/- 45 degrees) the
adapting orientation (an orinetation-selective
effect).
CS
Adaptation orientation
Orinetation
22
Role of spatial frequency-specific adaptation in
the preception of blur. Recall that most low
myopes do not see the world as blurry.
Fixate
During adaptation the face on the left has extra
high SFs, and thus creates extra adaptation at
high SF, and thus when we subsequently view a
normal face, it looks as though is has less
contrast in the high frequencies, which is the
same as blur. The converse is true for the left
image.
23
Explanation of contrast adaptation neural
fatigue! Neurons that respond repeatedly during
the adaptation period of the experiment become
fatigued and less able to respond. Thus when
previously visible low contrast test stimuli are
presented, they no longer elicit a response, and
thus we observe an elevated contrast threshold
(reduced CS). Important observation of
orientation and SF selectivity Shows that
neurons that detect stimuli of different
orientation or SF from the adapting stimulus were
NOT FATIGUED and thus did NOT RESPOND to the
adaptation stimulus. That is, the neurons
respond to a select range of SFs and orientations.
24
Masking is also SF and orientation selective.
Demonstration.
Test still visible when presented with high
contrast orthogonal mask
Low contrast, but clearly visible test grating
Test still visible when presented with high
contrast mask of very different SF
Test invisible when presented with high contrast
mask of similar SF and Orientation
25
Example of noise masking which range of Spatial
Frequencies is the most effective masker for
character recognition? What does this tell us
about which SF range is used to read letters?
26
Forward and Backward masking
Explanation due to visual persistence, responses
linger after stimulus is turned off, therefore,
although the stimuli are not present
simultaneously, some portion of their neural
signals will overlap in time if the test and mask
occur close together in time.
27
Lateral Masking Masking can also occur when the
mask and test and not spatially superimposed.
This type of masking is usually called lateral
masking. A common example of this is the well
known phenomenon of crowding in which letter
acuity is reduced by the presence of near-by
contours which can be bars or other letters.
Because of the crowding phenomenon letter acuity
charts endeavor to keep crowding constant at
either a low or high level.
For example, one letter acuity chart designed to
detect amblyopia surrounds the test letters with
other crowding letters because amblyopes are
particularly susceptible to crowding.
28
Masking by pixelation This was once used by TV
police shows until someone pointed out that by
simply blurring image or walking away from TV the
face became visible.
29
Schematic representation showing coding of
orientation and spatial frequency within an
ocular dominance column.
Spatial description of individual orientation and
SF selective RFs.
Position of cells in cortex
30
Schematic of LGN shows that alpha and beta cells
project to different layers of the LGN.
Significance the two cell types, which have
different physiological properties, are
segregated in the LGN, and their subsequent
projections into and around cortex differ.
Sustained What
Transient Where
31
Parallel Pathways evidence for a what vs. where
split
Color Shape
Motion Stereo
32
Flickering grating
Stationary grating Carefully fixated
CS
Is the increase in CS at low SF due to
recruitment of magno/transient cells, or due to
the increased responsivity of parvo/sustained
cells to stimulus transients?
Spatial Frequency
33
How does retinal image motion affect CS?
Normal CSF (careful fixation, stationary stimulus)
Drifting stimulus
Partial retinal stabilization
CS
(almost) complete retinal stabilization
Spatial Frequency
Conclude (1) all contrasts are virtually
invisible without retinal image motion (cannot
see shadows of retinal vessels) (2) slowing
retinal image motion makes low spatial
frequencies disappear (carefully fixated blurred
objects fade and disappear Troxler Fading) (3)
introducing fast retinal image motion makes high
SF disappear (try reading text on a moving page,
motion blur).
34
Clinical Relevance Patients with abnormal eye
movements will have abnormal retinal image
motion. For example, Nystagmus,
Amblyopia. Experimental studies find that visual
acuity with increased eye movements is not as bad
as one might expect. Explanation as long as
the eyes stop moving for a moment, a good image
is produced.
35
Local lesions in LGN produce localized visual
field deficits due to retinotopic mapping of
geniculo-striate pathway.
Present test stimuli in this field location
find that Parvo lesions produce major deficits in
color vision, visual acuity, contrast sensitivity
for stationary gratings and even low spatial
frequency CS for flickering gratings. With Magno
lesions, vision is almost normal, even for
flickering gratings.
These studies challenge the idea that the Magno
cells are responsible for low SF high temporal
frequency sensitivity. Raises the question of
just what do they do?
36
Dyslexia paradox theory is that dyslexics have
problem coding transient signals. However, they
end up with a reading/writing problem, and surely
reading text is a spatial task. How can a
spatial task deficit be blaimed on a temporal
processing deficit?
Evidence some indication that dyslexics have
reduced sensitivity to transient stimuli, but the
evidence is week and some studies find normal
temporal sensitivity in Dyslexics.
37
Hypothesis for role of Magno Cells Suppress
after-images in Parvo system.
The after-image problem each new fixation
produces two neural signals, one from the new
foveal retinal image, and one lingering
after-mage from the previous fixation.
Fix1
Fix2
Fix3
Neural response
time
fixations
Dt
Eye position
saccade
time
Note that the neural response is delayed
(latency) and lingers on (persistence) after the
stimulus has been removed (fixation over). The
neural persistence might be a positive or
negative after-image.
38
Problem of overlapping neural signals originating
from different fixations when reading.
Neural response
SecondFix
Fixateone
Fixateone
Finalword
SecondFix
Finalword
SecondFix
Fixateone
Clearly, we do not experience this. Does the
magno (fast and transient) system act as a neural
eraser removing (suppressing/inhibiting)
lingering neural signals from previous fixation?
39
ALL M-cells will respond at same time due to a
saccade (transient). They respond faster than
p-cell and thus, can inhibit neural image in
p-cell neurons left over from previous fixation
before new image arrives (neural eraser).
Neural response
SecondFix
Fixateone
Finalword
P-cell response
M-cell response
P-cell Response after inhibition by M-cell
40
Some illusions that remind us that the perception
process can make errors.
http//www.illusionworks.com/html/cafe_wall.html
41
(No Transcript)
42
(No Transcript)
43
How many black dots are there?
44
(No Transcript)
45
An excellent example of perceptual instability
generated by retinal image ambiguity. There are
two realities that can generate this same image.
Can you see them both?