What is Binocular Disparity?
Our two retinas are offset from each other and technically don’t see the exact same thing (they produce different images)
What is Binocular Summation?
Combining the signals from our two eyes, which allows for inferences that aren’t possible with just one, such as stereopsis
What is stereopsis?
The perception of depth as calculated by multiple viewpoints
What are depth cues?
Information that tells us about a 3rd dimension in space
Can be either monocular or binocular
What are the monocular depth cues?
Occlusion, Relative Size and Height, Texture Gradients, Familiar Size, Aerial Perspective, Linear Perspective, Motion Parallax, and Shadow
Occlusion
Infer the ordering of objects in 3D space based on which one ones are occluding others
Relative size
Generally assume that smaller items are further away
Relative Height
For objects that are touching the ground, we assume that the higher object in our visual field is further
Familiar Size
Relative size and height often rely on our previous knowledge of object characteristics
Texture Gradients
Textures become harder to resolve as further distances due to higher spatial frequency
Aerial Perspective
More distant objects appear hazier, less distinct, and more desaturated (white)
This is because light scatters more before it reaches us
Linear Perspective
Lines that are parallel to each other in a 3D world will appear to converge as they extend away from us, until they eventually reach the vanishing point (where they appear to meet)
Motion Parallax
Objects that are closer to us will appear to pass by more quickly than things that are further away
Shadow
Our brain makes assumptions about illuminants to determine how shadows should be cast
- Helps with depth as well as color constancy
- Is tied to colors changing in brightness (distance from black)
Why have many species evolved two eyes?
Triangulation - the distance of an object can be solved for trigonometrically by understanding the distance between 2 other points (in this case, 2 eyes) and the angle at which the object is in relation to those points
How does eye position/angle change with distance?
Convergence: eyes turn inwards to look at near objects
Divergence: eyes turn outwards to look at far objects
What is the horopter?
An imaginary arc in our vision where there is no binocular disparity, meaning that the image of the object lines up at corresponding retinal points (aka the same distance from the fovea)
The horopter includes our fixation point (where the corresponding point is the fovea) and moves around with it
What is the result of no binocular disparity in the horopter?
Single vision stereopsis, meaning we can still get a sense of depth from the multiple view points, but it produces a single image of a 3D object
What is Panum’s Fusion Area?
The zone around the horopter where binocular disparity is small enough that our brain automatically solves for it and gives us single vision
What is the result of binocular disparity beyond the horopter and panum’s fusion area?
Diplopia, or double vision
Can be crossed or uncrossed depending on whether the item is in front of or behind our fixation point
The degree of binocular disparity and direction of perceived shift can then be used as a depth cue
Crossed diplopia
Items are perceived to move right in the left eye, and left in the right eye (relative to the fixation point)
This occurs when items are CLOSER than the fixation point
Uncrossed diplopia
Items are perceived to move right in the right eye, and left in the left eye (relative to the fixation point)
This occurs when items are FURTHER than the fixation point
What is the correspondence problem? What is it affected by?
If there is is binocular disparity and the images do not line up on corresponding retinal points, our brain has to figure out how to match things up
This is affected by spatial frequency (lower frequencies are easier to correspond due to less fine details)
What heuristics (assumptions) constrain the way our brain solves the correspondence problem?
Uniqueness: any given feature should be represented exactly 1 time on each retinal image
Continuity: neighboring points that aren’t edges of objects should be at similar distances
