Cognitive Psychology |Lecture 5|22nd Oct
VISION
- Provides information about our environment
- One of our sense and window to outside world
- Turns meaningless 2D array of intensity into 3D dimensional, meaningful objects
(interpretation of low-level visual information)
SOME MAZING FACTS
- Interaction of low-level visual processes, high level visual processes and LTM
- Objects and background
- Movement
- Shape shifting
THEORIES OF VISUAL PERCEPTION
- Higher cognitive processes
- Top down: the flow from the brain/ high level processes towards the raw data, role
of knowledge, memory, expectations, increases speed of processing (constructivist
theories)
- Bottom up: the flow of raw data/physical input towards the brain, data driven (direct
perception theories)
DIRECT PERCEPTION |Gibson 1966,1979
- Bottom up theory
- The pattern of light which reaches the eye contains all the visual information from
the environment (an optic array)
- The optic array provides unambiguous/invariant information about the object in
space
- Limitation: can’t explain constancy, can’t explain optical illusions
CONSTRUCTIVTIVE THEORIES |Gregory 1972, 1980
- Top down theory
- Nearly exact opposite Gibson
- Perception is not a direct consequence of sensory input. It is the result of a ‘best
guess’ to identify the sensory input
The same physical stimulus can be interpreted differently
We use other cues in the situation to resolve ambiguities
Perception is totally conceptually driven, or top-down: role of context
X non of the theories is 100% correct X
, NEISSER CYCLIC MODEL |1964’s
Cycle of: sensory cues from environment-top
down search for expected features-perceptual
model-bottom up analysis of sensory features
This model runs constantly
OBJECT RECOGNITION
- How do we recognize object? Whole object? Assembled parts? Edges or surfaces?
Textures? Prototypes?
Object recognition: THEORIES
1. Template matching models (viewer-centered, normalization stage and matching)
Compare to templates in memory until a match is found
Problem of imperfect matches
Can’t account for flexibility of pattern recognition system
2. Feature analysis model (feature detection theory)
Every object is made out of features
Pandemonium theory:
3. Recognition by components (object-centered)
2
VISION
- Provides information about our environment
- One of our sense and window to outside world
- Turns meaningless 2D array of intensity into 3D dimensional, meaningful objects
(interpretation of low-level visual information)
SOME MAZING FACTS
- Interaction of low-level visual processes, high level visual processes and LTM
- Objects and background
- Movement
- Shape shifting
THEORIES OF VISUAL PERCEPTION
- Higher cognitive processes
- Top down: the flow from the brain/ high level processes towards the raw data, role
of knowledge, memory, expectations, increases speed of processing (constructivist
theories)
- Bottom up: the flow of raw data/physical input towards the brain, data driven (direct
perception theories)
DIRECT PERCEPTION |Gibson 1966,1979
- Bottom up theory
- The pattern of light which reaches the eye contains all the visual information from
the environment (an optic array)
- The optic array provides unambiguous/invariant information about the object in
space
- Limitation: can’t explain constancy, can’t explain optical illusions
CONSTRUCTIVTIVE THEORIES |Gregory 1972, 1980
- Top down theory
- Nearly exact opposite Gibson
- Perception is not a direct consequence of sensory input. It is the result of a ‘best
guess’ to identify the sensory input
The same physical stimulus can be interpreted differently
We use other cues in the situation to resolve ambiguities
Perception is totally conceptually driven, or top-down: role of context
X non of the theories is 100% correct X
, NEISSER CYCLIC MODEL |1964’s
Cycle of: sensory cues from environment-top
down search for expected features-perceptual
model-bottom up analysis of sensory features
This model runs constantly
OBJECT RECOGNITION
- How do we recognize object? Whole object? Assembled parts? Edges or surfaces?
Textures? Prototypes?
Object recognition: THEORIES
1. Template matching models (viewer-centered, normalization stage and matching)
Compare to templates in memory until a match is found
Problem of imperfect matches
Can’t account for flexibility of pattern recognition system
2. Feature analysis model (feature detection theory)
Every object is made out of features
Pandemonium theory:
3. Recognition by components (object-centered)
2
