6 -Depth Perception Flashcards
why is depth perception important?
3D perception is vital for interacting with the world and recognising objects.
why is depth perception a problem
How do we obtain a 3D
percept from two 2D
images on the retina?
what is the inverse problem
Any retinal image is consistent with infinitely many possible configurations of the world
what are binocular cues
– Binocular disparity (Stereo vision)
Two eyes looking at stimulus from different locations
Visual system matches left and right image
Different images - use binocular disparities differences to see what the depth looks like
what are motion cues
we move objects move
Things that are close to us will move more and things further away will move less
In the opposite direction to direction were travelling
motion parallax - due to self-motion
kinetic depth (KDE) - due to object motion
When an object is moving will cause motion in the retinal image
Different parts of surface will move at different speeds, texture closest moves fast but texture at side moving slower and less
what is kinetic depth
due to object motion
When an object is moving will cause motion in the retinal image
Different parts of surface will move at different speeds, texture closest moves fast but texture at side moving slower and less
what is motion parallax
due to self-motion
Things that are close to us will move more and things further away will move less
In the opposite direction to direction were travelling
what are the 6 pictorial cues
texture elevation relative size perspective shading occlusion
what is texture
change in density and size
higher density further away
what is relative size
big image closer
smaller image gurther away
what is perspective
lines converge as they get further away from us
what is shading
patterns of light and dark can give an impression of cavities and convexities
convergence for a far target and near target
Info we get bc of eyes
Converge towards each other to fixate on certain objects
Converge more when closer
Distance - parallel
sends to visul system how far away the thing were fixating on is
what is accomodation
Focused image on retina
Depending on how far away things is lens changes to focus light
Lens fat when thing close to you to bend image
Further away thinner stretched out more to bring far away object
Send signals to higher up in visual system to say how far away the thing is
If we have so many different cues to depth what is the problem?
- Many cues are ambiguous – 2D image compatible with infinite 3D worlds
- With multiple cues available – how do we perceive a single unified world?
ambiguity in depth perception
many of the 3D cues (particularly the pictorial
cues) are ambiguous
ambiguity in shading
This retinal image could have been created by: • a convex object that is lit from above (A) • a concave object lit from below (B) • (or it could be a flat surface with a pattern of light & dark greys painted on it)
how can we overcome ambiguity in depth perception
- Many of the 3D cues (particularly the pictorial cues) are ambiguous
- We can overcome this ambiguity by using prior knowledge, or prior assumptions, to interpret the image.
- These assumptions are gained through our knowledge and experience of the physical properties of the world.
• This is a type of top-down processing and supports the constructivist approach to vision
(see week 1)
e.g. interpreting train tracks
To interpret this image we can make the assumption that lines in the world tend to be parallel.
e.g. shading in image
To interpret this retinal image we use an assumption that light comes from above
convex chape
ambiguity in texture
This retinal image could have
been created by either:
1. A 2D square with a pattern of ellipses (ovals) that change shape
2. A 3D cylinder with a 3D pattern of circles wrapped around
to interpret ambiguity in texture
To interpret these retinal images we assume surface textures are isotropic (unbiased orientation) and homogeneous (uniform density)
• Making this assumption means any changes in image texture orientation or density are attributed to changes in 3D surface orientation
isotropic
no bias in its orientation
homogenous
surface texture have same elements
ambiguity in elevation
To interpret this image we assume that objects rest on a ground plane
what assumptions are we using
Perspective cue - lines converging
Make assumption lines in world parallel - converging in order for us to work out theyre further away
Lighter at top darker on bottoms - eg. Onions - light coming from above - convex
Texture on walls increasing in density further away - make assumption its isotopic and homogeneous
Elevation cue - towards closer further away higher - assumption resting on aa ground plane - higher up in image
Assumptions can lead to errors
Assumptions will be valid in most situations,
but in some cases they will not be valid, which will lead to perceptual errors (illusions)
what is an example of an erroneous assumption
Ames room
Assume is a regular shaped room - left side of scene is much further away
Assumption is invalid - experience an illusion - innacuarate perception cue
When we view the Ames room we assume lines are parallel or at right-angles, however this is not the case.
• The perspective assumption of parallel lines is invalid
which causes an erroneous percept
how to we overcome ambiguity in depyh perception
Many of the 3D cues (particularly the pictorial cues) are ambiguous
• We can overcome this ambiguity by using prior assumptions based on physical
properties of the world
• In situations where our assumptions are not valid we will make perceptual errors
multi-cue perception
Real-world scenes have multiple cues present
• Cues must be integrated to achieve a single unified percept
• Integration helps to overcome problems of:
– Reliability – Ambiguity – Conflict
integration of cues helps to overcome problems of …
reliability
ambiguity
conflict
what are the 3 types of integration
compromise
dominance
interaction
when multiple cues are available what happens
When multiple cues are available the visual system will try to integrate cues in order to obtain a unified percept
what happens in compromise
When two sources of depth information are conflicting, the brain will try to find a compromise between the two.
try to average
what did young et al do
Participants viewed computer generated cylinder defined by 2 cues in conflict:
– Texture – Motion
does the cylinder look flattened circular or stretched?
Change texture - stretched can see on side
Motion -
each define a different shape
texture - flattened
motion - stretched
texture and motion conflict - What shape do participants perceive?
Perceived shape is a compromise between the two cues
what if one cue is more reliable than the other
When texture made less regular (hence less reliable) perceived shape is biased towards motion cue
When motion made less smooth (hence less reliable) perceived shape is biased towards texture cue
summary of compromise
- When 2 cues are conflicting the brain will try to average them
- Final percept of shape will be biased towards most reliable cue
what happens in dominance of cues
when two cues define very different shapes or depths the brain may choose to ignore one in preference for the other
large cue-conflicts tend to arise when ones assumptions are invalid
usually the cue with valid assumptions will dominate
when an invalid cue dominates what happens
it creates an illusion
what happens in when cues are in conflict
When conflict between 2 cues is too great, the
brain will choose one cue and ignore the other
how to disambiguate texture
could be convex or concave
usually percieved as convex bc we have a bias for convex objects
binocular disparity can disambiguate
binocular disparity and texture interaction
Adams and Mamassian (2004)
introduced conflict between cues
•Texture disambiguated by binocular disparity (interaction)
then 2 cues averaged
what is the order of integration
interaction stage - cues disambiguate each other
compromise stage - join together
summary of interaction
- Ambiguous cues such as texture and shading can be disambiguated by other less ambiguous cues
- Evidence suggests this stage occurs prior to cue compromise.
what are the 3 types of cue integration
Three types of cue integration
- Compromise
- Dominance
- Interaction
Cue integration helps to overcome problems of:
– Unifying our percept of the world
– Differences in cue reliability (compromise)
– Conflicting information (compromise & dominance)
– Cue ambiguities (interaction)
Do cue conflicts apply in the real world?
yes Cue conflicts arise in a number of everyday stimuli e.g. – Art & Photography – Movies – 3D Movies
movies - cue conflicts
pictorial and motion cues specify a 3D scene, but binocular cues and oculomotor cues define a flat surface
3D movies
3D movies – pictorial, motion and binocular cues all specify 3D scene, only oculomotor cues specify flat surface
art and photography - cue conflicts
Art & Photography: binocular, motion and oculomotor cues will all tell you these are 2D surfaces and yet we perceive vivid three- dimensionality due to pictorial information
Depth cues must be integrated to:
– achieve a unified percept
– take account of differences in reliability
– overcome conflict and ambiguity
what is an inverse problem
recoving depth drom 2D retinal images
what are the 3 types of integration
compromise dominance interaction
how is ambiguity overcome
by the use of prior assumptions