Object Perception

Question 1

perception:

Answer 1

obtaining a description of shape, size, material composition, etc from light info

Question 2

Recognition

Answer 2

the matching of this description with something previously stored in memory (more top-down process)

Question 3

Edge detection:

Answer 3

Problems:
– Missing edges
– Spurious edges
– Integrating local edges into global contours
– Distinguishing different kinds of edges

•  Object boundary (depth discontinuities)
•  Surface markings
•  Shadows
•  texture

Question 4

Edge classification

Answer 4

Important edges are given by differences in:  
•  luminance 
•  color 
•  motion 
•  depth
 •  texture

Question 5

Junction detection and classification – relatability, non-accidental features.

Answer 5

Presence or absence of junctions determines important aspects of visual processing.

• T junctions: occluding edges
• X junctions: transparency edges
• L junctions: object corners
• Y junctions: corners of 3D objects

Question 6

Boundary assignment

Answer 6

-space has a large effect on what is perceived as one object vs several objects

• Unit Formation / Object Formation
• Segmentation and Grouping
• Classical approach:Gestalt principles

Question 7

Gestalt Theory

Answer 7

1)   Gestalt = overall form (put together, organized structure)
2)   1890-1930
3)   Perception of a thing (e.g., object, scene, music) is more than the sum of its parts:

Question 8

Gestalt Principles

Answer 8

•   Good continuation
•   Similarity
•   Proximity
•   Common fate
•   Closure
•   Symmetry
•   Simplicity or good figure

Question 9

Unit formation (Problems in object perceptions) – Gestalt Principles

Answer 9

Unit formation: visual system connects spatially separated visible areas using two processes:

contour interpolation & Surface interpolation

Question 10

Contour interpolation:

Answer 10

1. The process begins with the locating of contour junctions.
2. Interpolated edges begin and end at these junctions.
3. Contour interpolation follows a smoothness constraint, known as contour relatability
4. Relatability is related to the Gestalt idea of good continuation

Question 11

Surface interpolation:

Answer 11

1. The contour interpolation process depends on oriented edges leading into junctions.
2. The surface process complements contour interpolation.
3. Surface properties “spread” under occlusion within real and interpolated boundaries.
4. This process depends crucially on matches of color, lightness, and texture.

Question 12

Object Ambiguity: ambiguous figures and accidental viewpoints

Answer 12

• happens with occlusion

- accidental view: you have to be at a very specific position to see independent objects as connected vs as occluded

Question 13

Template Representation:

Answer 13

any pattern that would fit the template A would be recognized as letter A

• its specific to orientation in plane, depth, size, location, etc.
• issue: you would need too many templates

Question 14

Structural Representation:

Answer 14

volumetric representation

Examples: 
Polyhedra 
Superquadratics 
Generalized cylinders 
Geons

Question 15

Recognition by Components Theory:

Answer 15

• All objects are composed of a particular subset of geons in a particular spatial configuration
• Recognizing an object involves recognizing the particular geons involved and their particular spatial configuration

Benefits: viewpoint invariant
the particular viewpoint does not impact recognition. Size, orientation, rotation do not matter. Nor do color, texture, or luminance.

Problems with RBC
Doesn’t work for natural objects or natural scenes (e.g, trees, lakes, animals, etc)

Question 16

Geon Theory

Answer 16

• geon representation is
• • scale invariant (size can change)
• • position invariant
• • rotation (viewpoint) invariant**
• • • not entirely invariant from study with greebles
• accounts for many features of human object recognition
• theory falls through in explaining how we can discriminate objects within the same category
• • different chairs vary in scale but we can discriminate them
• • also how can one describe a bird in terms of geons

Question 17

Elastic Graph representation

Answer 17

• allows invariance to size, position, orientation in plane and some degree of viewpoint invariance
• graphs are not rigid
• accounts well for face recognition or recognition tasks that require holistic info
• biologically plausible

Question 18

Feature-based representation

Answer 18

• based on collection of features
• ignores topological info
• allows fast recognition
• doesn’t account for the ability to discriminate between a two featured sets in different orders