aka Depth Perception 3D Space Perception The flat retinal image problem How do we reconstruct 3Dspace from 2D image What information is available to support this process Interaction between ID: 225089
Download Presentation The PPT/PDF document "3D Space Perception" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
3D Space Perception
(aka Depth Perception)Slide2
3D Space Perception
The flat retinal image problem
:
How do we reconstruct 3D-space from 2D image?
What information is available to support
this process?
Interaction between
Perceived Size
and
Perceived Distance
(both depend upon “scaling”)Slide3
Size Constancy
Perceived size is not slavishly linked to retinal size; otherwise your car would appear to be smaller when observed at increasing distances.
Instead, perceived size tends to remain invariant across observation distance…a phenomenon known as
size constancy
.
Perceived size depends upon the psychological scaling of retinal size relative to perceived distance. Hence, size, distance and 3D visual perception are all based upon a more complex process known as
spatial scaling
.
Corollary
:
Perceived Speed = retinal velocity x scaled distanceSlide4
Failures of Size Constancy
The Moon IllusionSlide5
Failures of Size Constancy
The
Buechet
Chair
Click
here
for moreSlide6
The Ames Room
(Iowa State University at Ames, IA)
Failures of Size ConstancySlide7
3D-Depth Information “Cues”Slide8
Oculomotor
Information
State of Accommodation
State of
VergenceSlide9
Accommodation
In theory, the
efferent
signal driving the
ciliary
muscles (and/or
afferent
feedback from stretch sensors in the
ciliary
muscles) could be used by higher-order visual processes to help scale 3D space and/or visual distance.
There is little evidence to support this hypothetical role of accommodation.Slide10
Vergence
Eye Movements
Support for the role of efferent/afferent 3D information from
vergence
eye movements comes from:
“Tower Speed Illusion
”Slide11
Static Monocular
Sources of 3D Information
Occlusion
Familiar Size (Relative Size)
Texture Gradients
Linear Perspective
Aerial Perspective (Atmospheric extinction)
Shadow CastingSlide12
Occlusion
Near objects
block visual access to far objectsSlide13
Linear Perspective
Parallel lines on the
v
isual plane converge
t
oward the “vanishing
p
oint” with increasing
o
bservation distance
This law of projective
g
eometry provides a
s
trong cue about distance
a
nd 3D space.Slide14
Linear PerspectiveSlide15
Linear Perspective in the Service of ArtSlide16
Familiar Size/Relative Size
Objects of the same physical size
project different size retinal images depending upon the observation distance.
This
knowledge
and prior experience contribute to 3D space perception.
What is the height of this
s
culpture in feet?Slide17
Familiar Size Cue
Easter Island Sculpture
w
ithout Familiar Size Cue
Easter Island Sculpture
w
ith Familiar Size Cue
Novel objects can be psychologically scaled given visual references
o
f known size. For example…Slide18
Texture Gradients
An extended surface
w
ith uniform spatial
t
exture will project a retinal image with a non-uniform texture gradient that increases in spatial frequency as observation distance increases.Slide19
Aerial Perspective
Particulate matter in the atmosphere scatters light; reducing contrast and intensity of the retinal image.
The light from distant objects must pass through more atmosphere than the light from near objects.Slide20
Shadow Casting
Just as occlusion of objects serves as a powerful cue for depth…
occlusion of the illuminant
(sun) forms shadows which provide a powerful source of information for extracting 3D representations from a 2D retinal image.Slide21
Identify the Monocular Depth Cues
A Rainy Day in Paris
Gustav
Caillebotte
(1848-1894) Slide22
Identify the Monocular Depth Cues
A Rainy Day in Paris
Gustav
Caillebotte
(1848-1894)
Linear Perspective
Occlusion
Texture Gradient
Aerial Perspective
Shadow Casting
Familiar SizeSlide23
Dynamic Monocular
Sources of 3D Information
Motion Parallax
Relative Angular Velocity
Radial Expansion/Looming
Moving ShadowsSlide24
Motion Parallax
Motion parallax occurs when an observer fixates a point at intermediate distance and then rotates their head.
Objects in the distance appear to move WITH head motion; while objects closer than the fixation plane appear to move AGAINST the rotation of the head.
Motion Parallax and Dynamic
Shadow Casting
DemoSlide25
Optic Flow: Radial Expansion
Optic Flow and Driving
DemoSlide26
Delta Angular Velocity/Angular Size
Lecture Note
:
Need for improved Slow Moving Vehicle sign/Slide27
Binocular Depth PerceptionSlide28
Advantages of Binocularity
Redundancy (survival value)
Stereopsis
(Predators)
Large Field-of-View (Prey)
Binocular summation improves sensitivity
by √2 (
signal:noise
ratio sampling theory)
Binocular acuity better than monocular;
Same for CSF and many other functionsSlide29
Binocular Vision
(cont.)
Binocular rivalry
(Role of the “dominant” eye)
Autokinesis
phenomenonSlide30
Stereopsis
Ability to use binocular
retinal disparity
information to extract relative depth information from the retinal image pairs
Retinal “mismatch” can be used to reconstruct much of the missing 3
rd
dimension from the flat retinal imagesSlide31
Retinal Disparity
Understanding begins with a consideration of the geometry of the
horopterSlide32
Horopter
(Corresponding Retinal Images)
The
HOROPTER
is an imaginary
s
urface whose points are all at the
s
ame distance as the fixation point.
Points on the
horopter
project to
c
orresponding locations on the temporal and nasal retinas,respectively.These corresponding locationsexhibit zero retinal disparityi.e., D = d
temporal – dnasal = 0Slide33
“Crossed” and Uncrossed” Retinal Disparity
The
corresponding
locations for the
“closer”
green stimulus
exhibits
positive
retinal
disparity
D =
d
temporal
– dnasal > 0(or “crossed” disparity)The corresponding locations for the“farther” red stimulus exhibitsnegative retinal disparity D = dtemporal – dnasal < 0(or “uncrossed” disparity)Slide34
Depth
Recovery
by
Binocular
Cortical
CellsSlide35
Panum’s
Fusion AreaSlide36
Nativists v. Empiricists “Debate”
Nativist position
The CNS is capable of processing many environmental invariants at birth – giving rise to direct perception (e.g., James Gibson)
Empiricist position
Sensory information is too impoverished to explain perceptual experience without recourse to
knowledge
about the world; it is based upon “unconscious inference” (e.g., Bishop Berkeley)Slide37
Support for Nativism
Eleanor Gibson’s
Visual Cliff
Experiment
(and HRD replication studies)
Bela
Julesz’s
Random Dot Stereogram
paradigmSlide38
Random Dot
Stereograms
Bela
Julesz
Can retinal disparity yield perception of depth independent of knowledge about the nature of the world?
Nativist vs. Empiricist DebateSlide39
Red-Blue Anaglyph Technique
(black background)
Anaglyph glasses transmit RED and MAGENTA dots to the left eye; and, the BLUE and MAGENTA dots to the right eye.
Demo stimulus from USD’s PSYC 301 Lab