Perception - Depth and Visual Scene Analysis

Question 1

Why is difficult to design perceiving machines?

Answer 1

Visual inputs provide ambiguous information about 3D structures of the world

Image complexity makes it increasingly tricky for computers to organise visual scenes into distinct objects

Question 2

What are the types of cues to infer depth?

Answer 2

Monocular (or pictorial) cues

Binocular cues

Question 3

What are monocular cues?

Answer 3

Work with one eye

Question 4

What type of depth cue is relative height?

Answer 4

Monocular

Question 5

How does relative height help infer depth?

Answer 5

Vertical position of different objects in image

Where their bases are

Objects that are below the horizon and have their bases higher are typically perceived as being more distant

Question 6

What type of depth cue is relative size?

Answer 6

Monocular

Question 7

How does relative size help infer depth?

Answer 7

If two objects are of equal physical size, the more distant one will take up less of your field of view

Need to know relative size of objects when judging distance (top-down processing)

Question 8

What type of depth cue is occlusion?

Answer 8

Monocular

Question 9

How does occlusion help infer depth?

Answer 9

Closer objects will occlude further away ones

Question 10

What type of depth cue is linear perspective?

Answer 10

Monocular

Question 11

How does linear perspective help infer depth?

Answer 11

Parallel lines extending away from observer converge in the distance

Question 12

What type of depth cue is texture gradient?

Answer 12

Monocular

Question 13

How does texture gradient help infer depth?

Answer 13

Texture elements get smaller and more dense with distance

Foreshortening (circles become ovals) occurs when the surface is tilted away

Question 14

What type of depth cue is motion parallax?

Answer 14

Monocular

Question 15

How does motion parallax help infer depth?

Answer 15

Happens when we move or objects move

Closer = move faster

Distant = move slower

As we move, more distant objects will glide past us more slowly than nearer objects

Question 16

What type of depth cue is shadows and shading?

Answer 16

Monocular

Question 17

How do shadows and shading help infer depth?

Answer 17

Cast shadows can create strong perception of depth

Brightness of surface depends on its orientation with respect to the light source

In cases of ambiguity, brain often assumes light comes from above

Question 18

What are binocular cues?

Answer 18

Requires both eyes

Question 19

What type of depth cue is binocular disparity?

Answer 19

Binocular

Question 20

What does stereoscopic vision mean?

Answer 20

Our two eyes receive a slightly different image of the world

Question 21

How does binocular disparity help infer depth?

Answer 21

Creates a difference in image location of an object seen by left and right eyes

Size of disparity depends on object’s depth

Via stereopsis

Question 22

What does corresponding position mean?

Answer 22

If one retina slid on top of other, points would overlap

Question 23

What is the horopter

Answer 23

Set of points in space that project to corresponding positions in the two retinas (zero disparity)

Includes the fixation point

Question 24

What is crossed disparity?

Answer 24

Objects closer than the horopter

No corresponding positions

Would have to cross eyes to fixate on it

Image lies further to left from right eye’s viewpoint than from left eye’s viewpoint

Image lies further to right from left eye’s perspective

Question 25

What is uncrossed disparity?

Answer 25

Objects further away from horopter

Would have to uncross eyes to fixate on it

Image lies further to right from right eye’s perspective

Image lies further to left from left eye’s perspective

Question 26

What is structuralism?

Answer 26

Reductionist approach pioneered Wilhelm Wundt

Need to understanding fundamental building blocks to understand perception

Question 27

What does structuralism believe perception is?

Answer 27

Sum of “atoms” of sensation

Question 28

What does the Gestalt school believe?

Answer 28

Pushed back against structuralism

Wertheimer, Kohler, Koffta

Whole form or configuration greater than sum of its parts

Question 29

What are illusory contours?

Answer 29

Images where boundaries of objects seen when no boundaries actually there

Some images evoke a perception of edges in locations where there is no change in luminance of colour

Difficult to explain via structuralist approach

Question 30

What are Gestalt principles of perceptual organisation?

Answer 30

All manifestations of Law of Pragnanz (‘good figure’)

Any image can be interpreted in multiple ways but we try and come up with simplest and best solution

Proximity
Similarity
Common fate
Good continuation
Closure
Symmetry

Question 31

What is Gestalt’s principle of proximity?

Answer 31

Elements closer together in space are more likely to be grouped together

Question 32

What is Gestalt’s principle of similarity?

Answer 32

Similar things grouped together

Could be in terms of any basic characteristic

Question 33

What is Gestalt’s principle of common fate?

Answer 33

Things that move together group together

Motion is a powerful way of segmenting objects from background

Motion tends to work with spatial processing

Question 34

What is Gestalt’s principle of good continuation?

Answer 34

Group elements to form smooth continuing lines rather than abrupt or sharp angles

Helpful when objects occluding each other

Question 35

What is Gestalt’s principle of closure?

Answer 35

Prefer interpretations when continuous figures

Group elements to form complete figures, even if incomplete

Uses top-down processing

Question 36

What is Gestalt’s principle of symmetry?

Answer 36

Formed in groups that are balanced or symmetrical

Question 37

What are the strengths of Gestalt principles?

Answer 37

Correct about many things - perceptual objects not just a sum of their parts

Principles generally hold across a wide range of images

Question 38

What are the weaknesses of Gestalt principles?

Answer 38

Vague and imprecise - not always clear what is meant by a “good” and “simple” shape

No workable theory of how it happens (underlying neural mechanisms)