16-17. depth 1&2 Flashcards
Many to one representation meaning?
There are many different 3D scenes that can produce the exact same retinal image
- a big balloon far away
- a small balloon really close
- a medium balloon in the middle
Retinal size depends on…? (2 things)
- size of object
- distance to object
Ways to divide depth cues?
(3 levels of tree diagram)
- oculomotor vs retinal image
- monocular vs binocular (disparity)
- mono: static vs dynamic
Oculomotor cues meaning? 2 types? Problem?
- cues based on feedback from muscles controlling shape of lens and position of eyes
- neither are very precise or super helpful
- accommodation
- convergence
2 types of monocular cues (and meaning) ?
- only need one eye to provide depth information
- static (doesn’t need motion)
–> all info is in a picture - dynamic (requires motion)
3 types of static cues?
- position based
- size based
- lighting based
2 position based cues?
- partial occlusion
- relative height
Position of things relative to each other
4 size based cues?
- familiar size
- relative size
- texture gradients
- linear perspective
change in size (relative) gives you depth info
3 lighting based cues?
- atmospheric (aerial) perspective
- shading
- cast shadows
lighting info = depth info
3 dynamic cues?
- motion parallax
- optic flow
- deletion and accretion
Accommodation def? Type? (Problem?)
Oculomotor
- detect change in ciliary muscles to adjust lens to focus on the object
- VERY limited range (<2m)
- Even within that range, may still be very imprecise
Convergence def? Type? (Problem?)
Oculomotor
- detect how converged your eyes become to focus on an object
–> far away = not converged
–> super close = very converged - VERY limited range (<2m)
–> beyond that, the angle between your eyes changes very little
Partial occlusion def? Type of cue?
Monocular - static - position based
If an object partially covers another, you perceive the former (cover-er) as closer than the latter (covered)
*very reliable
Relative height** def? Type of cue?
Mono - static - position
An object that is closer to the horizon (or eye level) appears farther away
Think of the beach/inlet scene
Size-distance relation meaning?
- the farther away an object is, the smaller its retinal image
- RETINAL size decreases proportional to the increase in distance
Familiar size def? Type of cue?
mono - static - size
- use the retinal image size of a familiar object to judge our distance to it
- we know the retinal image size of familiar objects at familiar locations
Relative size def? Type of cue?
mono - static - size
- if several objects (ex. people) are all about the same physical size, we use RELATIVE size of retinal images to judge distances
- familiar objects are the same size, but some appear much bigger –> helps judge distance
Texture gradient def? Type of cue?
mono - static - size
- if surface variations or repeated elements of a surface are fairly regular in size and spacing, the retinal image size of these equal-size features decreases as their distance increases
Accelerated texture gradient meaning?
- Adjusting the spacing between repeated elements in a design
- creates illusion to make a building look longer/taller etc
Linear perspective def? Type of cue?
mono - static - size
- parallel lines appear to converge as they get farther away
Atmospheric (aerial) perspective def? Type of cue? (why?)
mono - static - lighting
- parts of the image that are REALLY far away tend to be hazier, blurrier, and a little bluer
- the farther the distance = more air between you and the object
–> more light scattered = objects are blurrier - only works with large distances
Shape from shading def? Type of cue?
mono - static - lighting
- gives information about relative depth / orientation of different parts of the object
- people very biased to assume light source comes from above
What implicit assumption do we make about shading/lighting? Why is it important?
That the light source comes from ABOVE!
- if an object is light on top and shaded on the bottom, (most common) we think its raised above the background
- if an object is shaded on top and light on the bottom, we think its indented
Cast shadows def? Type of cue?
mono - static - lighting
- depth can be signaled by the shadows cast by an object
- shadows can make you think an object is moving away or staying at the same distance but floating
Motion parallax def? Type of cue?
mono - dynamic
- objects appear to move in diff speeds and directions in the retinal image as an observer moves ACROSS a scene
- things CLOSER move FASTER (and FARTHER)
Motion parallax –> how does where you’re looking affect the direction the objects “move”? (Ahead vs middle)
If you look straight ahead: all objects move OPPOSITE the direction you move
If you FIXATE on an object in the middle…
- objects closer than fixation move OPPOSITE
- objects further from fixation move SAME direction
Optic flow def? Type of cue?
mono - dynamic
- the relative motions of objects and surfaces in the retinal image as the observer moves forward or backward through a scene
- things closest to you move away from the focus of expansion much more quickly
Focus of expansion def?
- The point toward which you’re moving when talking about optic flow
- everything moves outward from that point
Deletion and accretion def? Type of cue?
motion - dynamic
- deletion: objects are hidden as they move behind things
- accretion: objects are revealed as they emerge from behind things
Stereopsis meaning?
The vivid sense of depth arising from the visual system’s processing of the different retinal images in the two eyes
Binocular disparity def?
A depth cue based on differences in the relative positions of the retinal images of objects in the two eyes
Corresponding points def?
A point on the left retina and a point on the right retina that would align if they were superimposed
Exs:
- foveas are corresponding
- 2 spots that are both 4mm Left of fovea
Non corresponding points def?
Any 2 points on L+R retinas that wouldn’t exactly line up
Ex:
- L fovea and 4mm off R fovea
Horopter def?
An imaginary surface defined by the locations in a scene from which objects would project retinal images at corresponding points
3* types of disparity?
- crossed disparity
- uncrossed disparity
- zero disparity* (no disparity)
Crossed disparity def?
- an object that is closer to you than the horopter has crossed disparity
- to fixate on it, you’d have to converge your eyes more than they already are
Uncrossed disparity def?
- an object farther than the horopter has uncrossed disparity
- not crossed
Zero disparity def?
- an object on the horopter has zero disparity –> NO disparity
- this applies to the object you’re fixated on and any other objects on the horopter
Magnitude of disparity?
- an object closer to the horopter (but not on it) will have LESS disparity than an object farther from the horopter
- regardless of whether its crossed / uncrossed
–> (closer/farther to YOU from the horopter)
Correspondence problem meaning?
how does our brain know that the L retinal image retina was produced by the same object as the R retinal image?
problem of determining whether/which features in one ret. image correspond to the other
Stereogram def?
Two depictions of a scene that differ in the same way as an observer’s two retinal images of that scene would differ; an observer who simultaneously views one image with one eye and the other image with the other eye (as in a stereoscope) will see a combined image in depth
Random dot stereogram def? Important for?
- 2 grids of randomly arranged dots –> identical except for the displacement of a portion in one image relative to the other;
- view in stereoscope or anaglyph (superimposed)
- you’ll see a single image with the displaced portion in depth
- important to determine if object recognition happens before or after matching corresponding parts of retinal image (correspondence problem)
Answer to correspondence problem? (What happens first - corr. Match or object recog?)
- correspondence matching has to happen for binocular disparity
- RDS doesn’t contain any objects. But you perceive depth. Therefore, correspondence matching PRECEDES object recognition
How does correspondence matching happen in the brain? (2 assumptions)
- each feature matches only one other feature
- visual scenes tend to have smooth and continuous surfaces with relatively few abrupt changes in depth
Binocular cells – what are they?
- neural mechanism for binocular disparity
- neurons that respond to the stimulation of their receptive fields in both eyes simultaneously
Binocular cells – how do they work?
- receptive fields for binocular cells are at different corresponding or non-corresponding points
- different binocular points are tuned to different DISPARITIES (crossed, uncrossed, or zero)
–> tuned to different magnitudes of disparity
Binocular cells – what needs to happen for it to respond?
- both receptive fields (in both eyes) need to be stimulated)
- these receptive fields are most often non-corresponding
Binocular cells – where are they?
Have been found all over!
V1, V2, V3
Dorsal (MT) “where/how”
–> reaching/grasp depth perception
Ventral (V4/IT) “what”
–> depth for object shape perception
How early is depth information available in visual processing?
- very early!
- prior to visual search
- during pre-attentive processing
