Chapter 6: Depth Perception Flashcards
Cue Approach to Depth Perception
Compare images on retina
Stereoblind
Couldn’t experience vivid sense of depth that occurs when brain combine different retinal images of two eyes into single image
Many-to-one
Many different 3-D scenes can produce one and the same retinal image
Different Objects, Same Retinal Image
Can produce same retinal image if arranged appropriately
Depth Cues
Oculomotor Cues-gives info from muscles of eyes
Monocular Cues- static/ pictorial vs dynamic
Binocular Cues- compares one eye vs the other
Oculomotor Depth Cues
These cues arise from the workings of two different sets of oculomotor muscles
- Accommodation
- Convergence
Causes that are based on feedback from oculomotor muscles controlling the shape of lens and positions of eyes
Accommodation
How shape of lens adjusts to focus on image sharply
Convergence
Depth cue only over fairly short distance in front of eyes
Angles:
- 10 cm- 33 degrees
- 1 m- 3.5 degrees
- 4 m- less than 1 degree
Monocular Depth Cues
Cues that are based on retinal image and that provide info about depth even with only one eye open
- Static monocular cues (pictorial cues)
- involve motionless 2-D depictions of 3-D scenes
- include position, size, and lighting in retina
Position-based cues
- Partial Occlusion
- Relative Height
Partial Occlusion
Position-based depth cue- in scenes where one object partially hides another object, occlusion indicates that former is closer than ladder
- partial occlusion is an extremely common and reliable depth cue
- Intersection between two objects are called T-junctions
- Interposition- making predictions about unknown in space
Relative Height
- Allows inference of depth from position of objects relative to the horizon or eye level
- Provides info about the objects’ relative distance from the observer
- Affects depth perception even in absence of visible floor or ceiling
Size in the retinal image
What size-based cues tell us
- Size-distance relation
- Visual angle
- Size perspective
Size- distance relation
The further away on object is from retina, the smaller its retinal image
Visual Angle
Angle subtended (occupied) by an object in field of view
Size perspective
Depth info in scenes in which size-distance relation is apparent
Sized-based cues
Familiar size
Relative size
Texture gradient
Linear perspective
Familiar Size
Knowing retinal image size of familiar object at familiar distance lets us use its retinal image size to gauge its distance
- requires top-down processing
Relative Size
Under assumption that two or more are about same size, size of retinal image can be used to judge their relative distance
- same size in real life, but different on retinal images
Texture Gradient
If surface variations/ repeated elements of surface are fairly regular in size and spacing, the retinal image size of these equal-size features decrease as their distance increases
Linear Perspective
Parallel lines appear to converge as they recede in depth
Lighting-based cues
- Atmospheric perspective
- Shading
- Shadows
Atmospheric Perspective
The farther away an object is, the more air the light must pass through to reach us and the more light can be scattered, with result that distant objects appear less distinct than nearby object
Shading
Light falls on curved surfaces to create shading differences
*our natural assumption is that light comes from above
Shadows
Depth can also be signaled by the shadows cast by objects
Dynamic Cues
Movement in the retinal image
- still from one eye
- Motion-based cues:
- Motion parallax
- Optic flow
- Deletion and accretion
Motion Parallax
Dynamic depth cue that involves the different in the speed and direction with which objects appear to move in the retinal image as an observer moves within a scene
- more distant objects move more slowly across retina than nearby objects
Optic Flow
Dynamic depth cue that refer to the relative motions of objects and surfaces in the retinal image as the observer moves forward or backward through a scene
- focused on direction of movement
- importance of focus of expansion in optic flow
Focus of Expansion
Objects and surfaces near point toward which you’re heading move outward slowly in retinal image
- not moving/ static
- everything else around it moves
Deletion
Dynamic depth cue that refers to the gradual hiding (occlusion) of an object as it passes behind another one
Accretion
Dynamic depth cue that refers to the gradual revealing (“de-occlusion”) of an object as it emerges from behind another one
Binocular Disparity
Depth cue based on differences in relative positions of the retinal images object in the two eyes
- off by about 6 cm
Stereopsis
Vivid sense of depth arising from visual system’s processing of different retinal images in two eyes
Corresponding Points
Point on left retina and point on right retina were superimposed
Noncorresponding Points
Point on left retina and right retina that wouldn’t coincide if two retinas were superimposed
Corresponding and Noncorresponding Points
- Most of the images projected onto the retina will involve noncorresponding points
- The images will fall on different parts of the left and right retinas
Horopter
An imaginary surface defined by locations in a scene from which objects would project retinal images and corresponding points
- Corresponding points fall on horopter
- objects in front of or behind horopter land on noncorresponding points
Three types of binocular disparity
Crossed disparity- type of binocular disparity produced by an object that is closer than horopter
Uncrossed disparity- type of binocular disparity produced by object that is further than horopter
Zero disparity- type of binocular disparity where retinal image of object falls at corresponding points in two eyes
Disparity in the Retinal Images
- people are remarkably sensitive to very small binocular disparities
- stereopsis typically provides information about relative depth out to a distance of about 200m, which clearly provides a very useful adaptive, evolutionary advantage
Correspondence Problem
Problem of determining which features in retinal image in one eye correspond to which feature in retinal image in the other eye
Two hypotheses to solve the correspondence problem
- The visual system surveys the left and right retinal images and separately performs 2-D object recognition on them. The visual system in effect labels each feature of each retinal image as belonging to an object in the scene
- The visual system matches parts of the retinal images based on very simple properties such as color or edge orientation before proceeding to object recognition
Stereogram
Two depictions of scene that differ in same way as an observer’s two retinal images of that scene would differ
- Observer who simultaneously views one image with one eye and other image with other eye (in stereoscope) will see combined image
- Wheatstone was the first to explain stereoscopic vision and created the stereogram
- Person looking into the stereoscope sees a single image of the scene that gives a vivid impression of depth
Anaglyphs
superimposed images that are in blue and red and are set 6 cm apart from each other
Random Dot Stereogram (RDS)
Stereogram in which both images consist of a grid of randomly arranged black and white dots, identical except for the displacement of a portion in one image relative to the other
RDS provides a strong argument that […] must precede […]
RDS provides a strong argument that correspondence matching must precede object recognition
- Correspondence matching is necessary for perception of binocular disparity
- If object recognition precedes correspondence matching, RDS wouldn’t produce sense of depth, because RDS doesn’t have any objects
- Just random array of square dots or tiny scribbles, none of which can be separately labeled in either of two images - RDSs produce sense of depth, therefore correspondence matching must precede object recognition
How does the brain actually solve the correspondence problem?
- Each feature in one retinal image will match one and only one feature in other retinal image
- Visual scenes tend to consist of smooth and continuous surfaces with relatively few abrupt changes in depth
- Almost every point in field of view is surrounded by point at about same depth
Neural Basis of Stereopsis
Visual system contained neurons that respond to binocular information (and each of the disparity types) throughout the visual system
Binocular cells
Neurons that respond best to stimulation of receptive fields in both eyes simultaneously (sensitive to disparity)
- Found throughout visual pathways: V1, V2, and V3
- Dorsal pathway: MT and intraparietal sulcus
- Ventral pathway: V4 and IT cortex
What happens when multiple depth cues are present?
- The visual system relies on all information it receives to help determine depth and distance
- The redundancy is a good thing because determining depth and distance is such an important task
- Allows us to keep sense of depth
Research finding on integrating depth cues suggest several basic principles
- No single depth cue dominates in all situations, and no single cue is necessary in all situations
- Partial occlusion come closest to always being dominant - The more depth cues that are present in a scene, the greater is the likelihood that we’ll perceive the scene in depth and the greater is the accuracy and consistency of depth perception
- Depth cues differ in the kinds of information they provide
- Use this to construct more accurate view of layout - Depth perception based on multiple cues is a rapid, automatic process that occurs without conscious thought (unconscious inference)
Perceptual Constancy
When you perceive some property of object as constant despite changes in sensory info used to perceive that property
Depth and Perceptual Constancy
Size constancy
Size- distance invariance
Shape constancy
Shape- slant invariance
Size Constancy
Tendency to perceive an object’s size as a constant despite changes in size of object’s retinal image due to object’s changing distance from the observer
Size-distance invariance
Relation between perceived object and perceived distance
*Perceived size of object depends on perceived distance and vide versa
Emmert’s Law
Size-distance invariance of retinal afterimages
- Perceived size of an object depends on perceived distance and vice versa
Shape Constancy
Tendency to perceive an object’s shape as constant despite changes in shape of object’s retinal image due to object’s changing orientation
Shape-slant invariance
Relation between perceived shape and perceived slant
-perceived shape of object depends on its perceived slant and vice versa
Illusions of depth, size, and shape
- Forced perspective
- Ponzo illusion
- Ames room
- Moon illusion
- Tabletop illusion
Forced Perspective
Illusion in which a near and far object seem to be at about the same depth because of way they’ve aligned and the way they appear to be interacting, leading observer to disregard other depth cues in scene
Ponzo Illusion
Illustrates size-distance invariance and powerful influence of linear perspective on size- perception
- Activity in V1 (“bigger” object has more activity in V1)
Ames Room
-room specially designed to create an illusory perception of depth
- when viewed with one eye through peephole, objects along far wall look like they are all the same distance away, leading to misperception of their relative size
- can be seen as failure of shape constancy
Moon Illusion
Results from misperception of distance (closely related to Emmert’s law)
Tabletop Illusion
Visual system overcompensates edges
- Makes vertical table longer and horizontal table wider
3-D Motion Pictures and Television
-Polarized light- light in which wave oscillations are confined to particular direction
- Autostereoscopic techniques, including the parallax barrier method and a method using lenticular lenses, allowing 3-D viewing of displays without special glasses but are limited to one viewer or just a few viewers at a time
- Holograph creates 3-D images that can be seen in different perspectives from different locations
Parallax Barrier Method
Columns of pixels taken by left camera are given to left side and seen through slits (same thing on right side)
Lenticular Lenses
Lens strips, each strip conversing two columns of pixels
