1. INTRODUCTION
1.1. WHY IS VISUAL PERCEPTION COMPLEX?
1.1.1. OBJECTS OVERLAP
1.1.2. OBJECTS VARY WITHIN CATEGORY
1.1.3. ORIENTATION
1.2. WE IDENTIFY BUT ALSO ESTIMATE HOW OBJECT WOULD LOOK FROM DIFFERENT ANGLES AND ITS FUNCTIONS
2. PATTERN RECOGNITION
2.1. FEATURE-BASED APPROACH
2.2. MANY THEORIES ASSUME PROCESSING OF SPECIFIC FEATURES IS FOLLOWED BY GLOBAL PROCESSING BUT NAVON DISAGREES
2.2.1. LIMITATIONS OF NAVONS APPROACH
2.2.1.1. ATTENTIONAL PROCESSES INFLUENCE PERFORMANCE
2.2.1.2. FAILED TO DISTINGUISH BETWEEN ENCODING (NEURAL RESPONSES) AND DECODING (CONSCIOUS PERCEPTION)
2.2.1.2.1. MORE APPLICABLE TO DECODING
2.3. FEATURE DETECTORS
2.3.1. HUBEL AND WIESEL
2.3.1.1. A HIERARCHICAL VISUAL SYSTEM IN WHICH MORE COMPLEX FEATURES ARE BUILT (BOTTOM-UP) FROM SIMPLE ONES
2.3.1.1.1. DISCOVERED IN V1...
2.3.1.1.2. LIMITATIONS
3. PERCEPTUAL ORGANISATION
3.1. MANY OBJECTS OVERLAP MAKING PERCEPTUAL ORGANISATION HARD
3.2. GESTALTISTS
3.2.1. LAW OF PRAGNANZ
3.2.1.1. WE TYPICALLY PERCEIVE THE SIMPLEST POSSIBLE ORGANISATION OF THE VISUAL FIELD
3.2.1.1.1. LAW OF PROXIMITY
3.2.1.1.2. LAW OF SIMILARITY
3.2.1.1.3. LAW OF GOOD CONTINUATION
3.2.1.1.4. LAW OF CLOSURE
3.2.1.2. FIGURE-GROUND SEGMENTATION
3.2.1.2.1. FIGURE HAS A SHAPE IS IN FRONT OF THE GROUND AND CONVEX REGIONS BELONG TO IT
3.2.1.3. PERCEPTUAL ORGANISATION IS INNATE
3.2.1.4. FINDINGS
3.2.1.4.1. GEISLER WITH NATURAL STIMULI
3.2.1.4.2. COLOR LUMINANCE VS INFLUENCE THE PERCEPTION OF BOUNDARIES
3.2.1.4.3. UNIFORM CONNECTEDNESS
3.2.1.4.4. DISSIMILARITY
3.2.1.4.5. SOME PRINCIPLES ARE SHOWN AT VERY YOUNG AGE BUT SOME AREN'T SO LEARNING IS IMPORTANT
3.2.1.4.6. ATTENTION IS REQUIRED FOR GROUPING BY SHAPE SIMILARITY
3.2.1.4.7. FIGURE-GROUND SEGMENTATION CAN DEPEND ON EXPERIENCE AND MEMORY
3.2.1.4.8. BAYESIAN HIERARCHICAL GROUPING MODEL BY FROYEN
3.2.1.5. EVALUATION
3.2.1.5.1. FOCUSED ON KEY ISSUES
3.2.1.5.2. APPLICABLE TO NATURAL SCENES
3.2.1.5.3. MAIN IDEA IS FRUITFUL
3.2.1.5.4. THEY DE-EMPHASIZED ROLE OF PAST EXPERIENCE AND LEARNING
3.2.1.5.5. PRODUCED DESCRIPTIONS NOT EXPLANATIONS
3.2.1.5.6. MOST EVIDENCE BASED ON 2D DRAWINGS NOT REAL WORLD
3.2.1.5.7. THEY DID NOT DISCOVER ALL PRINCIPLES
3.2.1.5.8. FOCUSED ON ONE LAW AT A TIME
3.2.1.5.9. TOO INFLEXIBLE
4. APPROACHES TO OBJECT RECOGNITION
4.1. IDENTIFYING OBJECTS IN THE VISUAL FIELD
4.2. PERCEPTION-ACTION MODEL
4.2.1. MILNER AND GOODALE ARGUED OBJECT RECOGNITION AND PERCEPTION DEPEND PRIMARLY ON VENTRAL STREAM
4.2.1.1. THIS STREAM IS HIERARCHICALLY ORGANISED
4.2.1.1.1. RETINA
4.2.2. IT ACTUALLY INVOLVES NUMEROUS INTERACTIONS BETWEEN VENTRAL AND DORSAL STREAMS
4.3. SPATIAL FREQUENCY
4.3.1. VISUAL PERCEPTION SEEMS INSTANTANEOUS BUT IT IS NOT
4.3.2. COARSE-TO-FINE
4.3.2.1. SOME CELLS IN V1 RESPOND TO HSF AND CAPTURE FINE DETAIL
4.3.2.1.1. HSF RELATES TO COLUR SHAPE ETC.
4.3.2.1.2. CONSCIOUS PERCEPTION IS BASED ON INTEGRATED HIGH AND LOW SPATIAL FREQUENCIES
4.3.2.2. SOME CELLS RESPOND TO LSF AND CAPTURE COARSE INFO
4.3.2.2.1. LSF RELATES TO MOTION OR SPATIAL INFO
4.3.3. FINDINGS
4.3.3.1. VISUAL PROCESSING OF NATURAL SCENES IS COARSE-TO-FINE
4.3.3.2. GLOBAL PROCESSING OFTEN PRECEDES LOCAL PROCESSING BUT USE OF LSF AND HSF IS FLEXIBLE AND DEPENDS ON TASK DEMANDS
4.3.3.3. PUR CENTRAL VISION IS DOMINATE BY HSF WHILE PERIPHERIAL BY LSF MONA LISA'S SMILE IS LESS APPARENT IF FIXATED ON THE MOUTH
4.4. MARR'S COMPUTATIONAL APPROACH
4.4.1. OBSERVERS CONSTRUCT VARIOUS REPRESENTATIONS
4.4.1.1. PRIMAL SKETCH
4.4.1.1.1. 2D
4.4.1.1.2. DESCRIPTION OF MAIN LIGHT INTENSITY CHANGES
4.4.1.1.3. EDGES AND CONTOURS
4.4.1.1.4. VIEWPOINT-DEPENDENT
4.4.1.2. 21/2 SKETCH
4.4.1.2.1. DESCRIPTIONS OF DEPTH AND ORIENTATION OF VISIBLE SURFACES
4.4.1.2.2. VIEWPOINT-DEPENDENT
4.4.1.3. 3D MODEL REPRESENTATION
4.4.1.3.1. DECRIBES OBJECTS SHAPES AND THEIR RELATIVE POSITIONS
4.4.1.3.2. VIEWPPOINT-INVARIANT
4.4.2. EVALUATION
4.4.2.1. REALISED COMPLEXITY
4.4.2.2. SUCCESFULLY COMBINED IDEAS FROM NEUROPSYCHOLOGHY ANATOMY AND COMPUTER VISION
4.4.2.3. VIEWPOINT TRIGGERED OTHER RESEARCH
4.4.2.4. FOCUSED EXCESSIVELY ON BOTTOM-UP
4.4.2.5. SAID VISION ALWAYS TELLS THE TRUTH
4.4.2.6. PROPOSED PROCESSES ARE TOO COMPLEX COMPUTATIONALLY
4.5. BIEDERMANN'S RECOGNITION-BY-COMPONENTS THEORY
4.5.1. OBJECTS CONSIST OF BASIC SHAPES CALLED (36) GEONS
4.5.2. EDGE EXTRACTION
4.5.2.1. NON-ACCIDENTAL (INVARIANT) IMAGE PROPERTIES
4.5.2.2. VIEWPOINT-INVARIANT
4.5.2.3. DIFFICULT WHEN SOME GEONS ARE HIDDEN
4.5.2.4. STILL CAN RECOGNISE IF SUBOPTIMAL CONDITIONS
4.5.2.4.1. NON-ACCIDENTAL PROPERTIES
4.5.2.4.2. CONCAVITIES
4.5.2.4.3. REDUNDANT INFO
4.5.3. FINDINGS
4.5.3.1. NON-ACCIDENTAL PROPERTIES PLAY A VITAL ROLE
4.5.3.2. OBJECTS CONTOUR IS IMPORTANT
4.5.3.2.1. ESPECIALLY CONVEXITIES FOR SHAPE
4.5.3.3. CONTRARY TO THEORY OBJECT RECOGNITION NOT ONL DEPENDS ON EDGE BUT ALSO SURFACE
4.5.3.4. OBJECT RECOGNITION IS NOT AS VIEWPOINT-INVARIANT WITH UNFAMILIAR OBJECTS
4.5.4. EVALUATION
4.5.4.1. ROLE OF NON-ACCIDENTAL PROPERTIES AND GEONS
4.5.4.2. LIMITATIONS
4.5.4.2.1. FOCUSES ON BOTTOM-UP PROCESSES DE-EMPHASISES EXPECTATION
4.5.4.2.2. ACCOUNTS ONLY FOR UNSUBTLE PERCEPTUAL DISCRIMINATIONS NOT FOR FINE-GRAINED
4.5.4.2.3. INFLEXIBILITY
4.6. DOES VIEWPOINT INFLUENCE?
4.6.1. NO WHEN CATEGORISATION
4.6.1.1. BUT INITIAL PROCESSING IS VIEWPOINT.DEPENDENT
4.6.2. YES WHEN IDENTIFICATION
4.6.3. NEITHER YES NOR NO BECAUSE MULTIPLE KINDS OF INFO ARE USED IN FLEXIBLE MANNER TO ADAPT TO CONTEXT AND TASK DEPENDS ALSO IN BRAIN AREA
5. TOP-DOWN PROCESSES
5.1. MARR AND BIDERMANN EMPHASISED BOTTM-UP PROCESSES FROM VISUAL CORTEX TO INFEROTEMPORAL CORTEX BUT THIS IS OVERSIMPLIFIED
5.1.1. THERE ARE AS MANY BACKWARD PROCESSING NEURONS
5.1.1.1. RECURRENT PROCESSING IS OFTEN NECESSARY FOR CONSCIOUS VISUAL PERCEPTION
5.1.1.1.1. TOP-DOWN ESPECIALLY IMPORTANT WHEN BOTTOM-UP IS UNINFORMATIVE
5.2. PRIMES BIASED THE INTERPRETATION OF AMBIGUOUS FIGURES
5.3. CONTROVERSY
5.3.1. TOP-DOWN INFLUENCE RESPONSE BIAS ATTENTION OR MEMORY BUT NOT PERCEPTION
5.3.1.1. BUT ATTENTION OERCEPTION AND MEMORY ARE NOT CLEARLY SEPERABLE
5.4. VISUAL HALLUCINATIONS
5.5. VISUAL IMAGERY
5.6. BAR'S MODEL
5.6.1. OBJECT RECOGNITION DEPENDS ON TOP-DOWN PROCESSESING THE ORBITOFRONTAL CORTEX AND BOTTOM-UP PROCESSING IN THE VENTRAL VISUAL STREAM
5.6.1.1. ORBITOFRONTAL CORTEX USES CONTEXTUAL INFO AND FORM HYPOTHESIS ANDRESOLVE COMPETITION AMONG HYPOTHESIS
5.7. INTERACTIVE-ITERATIVE FRAMEWORK
5.7.1. OBSERVERS FORM HYPOTHESIS BASED ON THEIR GOALS KNOWLEDGE AND CONTEXT BEFORE PRESENTATION OF OBJECT
5.7.2. THEN RESOLVE COMPETITION AMONG HYPOTHESIS BY A DISTINGUISHING FEATURE OF OBJECT
5.7.2.1. ELEPHANT-TRUNK
5.8. TOP-DOWN PROCESSES INFLUENCE THE ALLOCATION OF LOCATION AND THIS INFLUENCES SUBSEQUENT BOTTOM-UP PROCESSING
6. FACE RECOGNITION
6.1. FACE VS OBJECT RECOGNITION
6.1.1. FACE MORE HOLISTIC PROCESSING (INTEGRATION OF FEATURES)
6.1.1.1. FASTER BECAUSE IN PARALLEL
6.1.1.2. MORE RELIABLE BECAUSE FEATURES CAN CHANGE
6.1.1.3. FACE-INVERSION EFFECT
6.1.1.3.1. EXPERIENCE FINDINGS ARE MIXED. YES FOR CARS NO FOR HORSES
6.1.1.4. PART-WHOLE EFFECT
6.1.1.5. COMPOSITE FACES
6.1.1.6. EVEN IN LITTLE CHILDREN
6.2. PROSOPAGNOSIA
6.2.1. BRAIN-DAMAGED PATIENTS WITH IMPAIRED FACE PROCESSING (FACE BLINDNESS)
6.2.1.1. ACQUIRED PROSOPAGNOSIA
6.2.1.1.1. CAUSED BY BRAIN DAMAGE
6.2.1.2. DEVELOPMENTAL PROSOPAGNOSIA
6.2.1.2.1. IN ABSENCE OF BRAIN DAMAGE
6.2.2. OBJECT AGNOSIA VS PROSOPAGNOSIA
6.2.2.1. THERE IS A DOUBLE DISSOCIATION
6.2.2.1.1. MORE PEOPLE HAVE IMPAIRED FACE RECOGNITION BUT INTACT OBJECT RECOGNITION
6.2.2.2. FINDINGS
6.2.3. FINDINGS
6.2.3.1. MANY PROSOPAGNOSICS SHOW COVERT RECOGNITION
6.2.3.1.1. FAMOUS FACES PRODUCE ACTIVITY
6.2.3.2. MANY PROSOPAGNOSICS HAVE FACE-SPECIFIC IMPAIRMENTS SO FACE RECOGNITION INVOLVES SPECIAL PROCESSES
6.3. FUSIFORM FACE AREA (IN VENTRAL TEMPORAL CORTEX)
6.4. EXPERTISE HYPOTHESIS
6.4.1. FUSIFORM EXPERTISE AREA?
6.4.1.1. EXPERTISE LEADS TO GREATER HOLISTIC PROCESSING
6.4.1.1.1. HOLISTIC PROCESSING IS NOT UNIQUE TO FACES BUT TO ANY OBJECT OF EXPERTISE
6.4.1.1.2. FUSIFORM FACE AREA MUST BE HIGHLY ACTIVATED WHEN CATEGORY OF EXPERTISE
6.4.1.1.3. PROCESSING OF FACES AND OBJECTS OF EXPERTISE SHOULD INTERFERE
6.4.1.2. ACCORDING TO THIS HYPOTHESIS WE ARE FACE EXPERTS BUT NOT TRUE FOR UNFAMILIAR FACES
6.5. BRUCE AND YOUNG'S THEOREOTICAL APPROACH
6.5.1. STRUCTURAL ENCODING
6.5.1.1. PRODUCES VARIOUS DESCRIPTIONS OF FACES
6.5.2. EXPRESSION ANALYSIS
6.5.2.1. PEOPLES EMOTIONAL STATES ARE INFERRED FROM THEIR EXPRESSIONS
6.5.3. FACIAL SPEECH ANALYSIS
6.5.3.1. SPEECH PERCEPTION IS ASSISTED BY LIP READING
6.5.4. DIRECT VISUAL PROCESSING
6.5.4.1. SPECIFIC FACIAL INFO IS PROCESSED SELECTIVELY
6.5.5. FACE RECOGNITION UNITS
6.5.5.1. CONTAIN STRUCTURAL INFO ABOUT KNOWN FACES
6.5.6. PERSON IDENTITY NODES
6.5.6.1. PROVIDE INFO ABOUT INDIVIDUALS
6.5.7. NAME GENERATION
6.5.7.1. PERSON'S NAME IS STORED SEPERATELY
6.5.8. COGNITIVE SYSTEM
6.5.8.1. CONTAINS ADDITIONAL INFORMATION AND INFLUENCES WHICH COMPONENTS RECEIVE ATTENTION
6.5.9. FOUR PREDICTIONS FOLLOW
6.5.9.1. THERE SHOULD BE MAJOR DIFFERENCES IN PROCESSING OF FAMILIAR AND UNFAMILIAR FACES
6.5.9.1.1. WE HAVE MORE STRUCTURAL INFO ABOUT FAMILIAR FACES
6.5.9.1.2. THERE IS WITHIN-PERSON VARIABLITY
6.5.9.2. SEPARATE PROCESSING ROUTES FOR FACIAL EXPRESSION AND FACIAL IDENTITY
6.5.9.2.1. YES BUT NOT ENTIRELY DIFFERENT JUDGEMENT ABOUT EXPRESSION IS INFLUENCED BY IDENTITY INFO AND VICE VERSA
6.5.9.3. WHEN WE SEE A FACE FAMILIARITY INFO FROM THE FACE RECOGNITION SHOULD BE ACCESSED FIRST
6.5.9.3.1. A FACE SHOULD NEVER LEAD TO RECALL NAME IN ABSENCE OF OTHER INFO
6.5.9.4. IMPAIRMENTS DEPEND ON STAGE
6.5.9.4.1. EARLY STAGES
6.5.9.4.2. FACE RECOGNITION UNITS
6.5.9.4.3. PERSON IDENTITY NODES BIOGRAPHICAL INFO
6.5.10. APPLIED TO DEVELOPMENTAL PROSOPAGNOSIA
6.5.10.1. SHOW RELATIVELY INTACT EARLY STAGES BUT LESS EFFICIENT LATER
6.6. SUPER RECOGNISERS
7. VISUAL IMAGERY
7.1. BETWEEN IMAGERY AND PERCEPTION
7.2. APHANTASIA
7.2.1. LACK OF VISUAL IMAGERY
7.3. WE ARE AWARE OF CONSTRUCTING THEM
7.4. HALLUCINATIONS
7.4.1. IF IMAGERY AND PERCEPTION CONFUSED
7.4.1.1. SCHIZOPHRENICS HAVE PROBLEM IN V1
7.4.1.1.1. DISTORTIONS IN TOP-DOWN PROCESSING
7.4.1.2. ANTON'S SYNDROME (BLINDNESS DENIAL)
7.5. WHY IMPORTANT?
7.5.1. ALLOWS TO PREDICT WHAT IF...
7.6. IMAGERY THEORIES
7.6.1. KOSSLYN
7.6.1.1. IMAGERY RESEMBLES PERCEPTION
7.6.1.1.1. THERE SHOULD BE FACILITATION AND INTERFERENCE
7.6.1.1.2. MUCH PROCESSING OCCURS IN EARLY VISUAL CORTEX
7.6.1.2. VISUAL IMAGES ARE DEPICTIVE REPRESENTATIONS
7.6.1.2.1. DISTANCE IN REPRESENTATION CORRESPONDS TO ACTUAL DISTANCE
7.6.1.2.2. THEY ARE FORMED IN VISUAL BUFFER (A SHORT TERM STORE AND IS FLEXIBLE)
7.6.1.3. VISUAL PERCEPTION INVOLVES BOTTOM-UP BUT IMAGERY TOP-DOWN
7.6.2. PYLYSHN
7.6.2.1. IMAGERY DOES NOT RESEMBLE PERCEPTION
7.6.2.1.1. INFO WITHIN IMAGES IS LESS DETAILED AND FLEXIBLE
7.6.2.1.2. OBJECTS BEING PERCEIVED OR IMAGINED ARE IDENTIFIED WITH ABOVE-CHANCE ACCURACY EXCEPT FOR IMAGINED OBJECTS IN V1
7.6.2.1.3. IDENTIFYING PERCEIVED OBJECTS IS GREATER IN AREAS OF EARLY VISUAL PROESSING AND OPPOSITE FOR IMAGERY
7.6.2.1.4. DJIKSTAR'S FINDINGS
7.6.2.1.5. FOR PERCEPTION STRONGER CONNECTIONS FROM V1 TO OTHER AREAS BUT WEAKER FROM HIGHER BRAIN AREAS BACK TO V1 AND OPPOSITE FPR IMAGERY
7.6.2.2. IMAGERY DOESN'T INVOLVE DEPICTIVE REPRESENTATIONS BUT TACIT KNOWLEDGE
7.7. BRAIN DAMAGE
7.7.1. IF THEY INVOLVE SAME MECHANISM BRAIN DAMAGE MUST HAVE COMPARABLE EFFECT ON BOTH
7.7.1.1. BUT THERE ARE MANY EXCEPTIONS
7.7.2. BRAIN AREAS INVOLVED IN EARLY VISUAL PROCESSING HAVE MORE EFFECT ON PERCEPTION BUT BRAIN AREAS ASSOCIATED WITH STORAGE OF INFORMATION HAVE MORE EFFECT ON IMAGERY
7.8. FINDINGS
7.8.1. ARE MORE CONSISTEN WITH KOSSLYN THAN PYLYSHYN
7.8.2. BUT THEY ARE LESS SIMILAR THAN SUGGESTED BY KOSSLYN