Midterm 2 Review Flashcards
(199 cards)
1
Q
How do we see colour?
A
Prism decomposes white light into the colour spectrum.
2
Q
Equation for light from surface
A
Illumination x reflectance
3
Q
Differences in colour vision example
A
Train painted “improved engine green” but the colour is red
4
Q
Colour vision deficiency
A
Red-green “colour-blindness”
5
Q
Why is red-green colour blindness more common in men?
A
It’s an x-linked trait so it’s on the X chromosome
6
Q
What are the three types of cone photoreceptors?
A
Short, medium, long
7
Q
Univariance
A
For one receptor, different combinations of wavelengths and intensity will produce the same response
8
Q
Why do we see in black and white at night?
A
You cannot perceive colour with only one receptor
9
Q
Why three cones?
A
Each cone by itself is colourblind, the combination of cones gives us colour perception.
10
Q
Does every species have three cones?
A
No, the number of cone types varies across species
11
Q
Tetrachromats
A
Birds and bees have four or more cone types - they have extra UV photoreceptors and can see more wavelengths.
12
Q
Trichromats
A
Such as humans - have three cones
13
Q
Dichromats
A
Many mammals such as dogs - have two cone types so can distinguish yellow from blue but not red from yellow.
14
Q
Missing cones
A
Most often missing medium or long cone types which causes red-green colour deficiency
15
Q
Red-green colour deficiency
A
Red and green are difficult to distinguish- when colour blind, colours are still perceived but difficult to distinguish.
16
Q
Cortical achromatopsia
A
Colour vision loss at a cortical level despite normal cone function - true colour loss
17
Q
Low pressure sodium lamp
A
Colour groups look different under a sodium lamp and are hard to distinguish - displays cortical achromatopsia
18
Q
Subtractive colour mixing
A
Light is subtracted by adding pigment because pigment absorbs light.
Start with white light such as in paintings
Red + green = brown
19
Q
Additive colour mixing
A
Add wavelengths of light to a surface that had no light
Start without light such as TVs and iPads
Combination of lights add together to produce colour
Green + red = yellow
20
Q
Thomas Young
A
Our eyes aren’t big enough to receive all wavelengths so maybe we have receptors which combine them (primary colours)
21
Q
Colour matching experiment
A
2 primaries aren’t enough but 4 is too many.
22
Q
Ewald Hering
A
Made the opponent colour theory to oppose the trichromatic theory
23
Q
Trichromacy in the eye
A
Wavelength is compressed into 3 dimensions (3 types of cone photoreceptor)
Perceived colour depends on relative strength of activation
24
Q
Evidence for trichromacy
A
Colour matching experiment -
People have to adjust the lights to match the colour provided.
25
Consequences of trichromacy
Multiple spectra can elicit same ratio of cones - thus appearing identical.
. Some colours are the same wavelength so they appear the same.
26
Opponent colour theory
Says that there’s 4 primary colours
- red, green, yellow and blue
These are organised into opponent pairs
- red-green and yellow-blue
27
Unique hues
Certain colour combinations don’t exist
- we can have reddish-orange and bluish-green but not a reddish-green or a bluish-yellow
28
Hue cancellation experiments
Adjust red light to cancel out green
- red-green combination seen as yellow
Adjust blue light to cancel out yellow
- blue-yellow perceived as white
29
Negative afterimages
Lilac chaser - see a green dot filling in the space but it’s not actually green.
30
Trichromacy vs opponency
Both can be correct
31
Two-stage model of colour coding
Different cones receive wavelengths that create different colours
32
What is the visible light spectrum?
400 - 700 nm
33
what are the three receptor types?
Short, medium, long
34
short cones
have peak sensitivity to short wavelengths
35
medium cones
have peak sensitivity to medium wavelengths
36
long cones
Peak sensitivity to long wavelength
37
principe of univariance
having a single cone type means you can’t discriminate between different wavelengths or intensities.
38
opponency
4 colour primaries organised in opponent pairs
- red green
- blue yellow
39
unique hues
Evidence for opponency:
- no such thing as a reddish green or a bluish yellow
40
colour cancellation
add green to “cancel” out red
- you get a yellowish brown colour instead of reddish green
41
negative afterimages
lilac chaser - see a green dot where the purple is not, but there is no green
42
retinal ganglion cells
excitatory and inhibitory receptive fields which react to different wavelengths.
43
white light has what wavelengths?
has short, medium and long wavelengths
44
neurons
react differently to different colours with different wavelengths.
45
chromatic edge detection
differentiate red and green etc as well as light and dark
- important for colour constancy
46
colour constancy
if the lights change (natural to fluorescent to dim to candle light etc) the colour of an object stays the same once you have seen it.
47
assumptions about colour
“Paint” versus “light”
Is it actually that colour or is shadow falling on it changing the way it looks?
48
cues for shadows
fuzzy edges and darker
The same shade could be perceived as darker if the brain thinks it’s a shadow
49
correspondence problem
when objects are moving, which goes with which?
- population coding helps to solve the correspondence problem (cues such as distance etc)
- are dots moving side to side or up and down?
50
local motion detectors
in the V1
- aperture problem: looking at movement through a small peephole means that the motion could be going any direction.
51
Global motion detectors
in MT
- get more motion information
-reicard detection (add more apertures to see flow of motion)
- MST optic flow
52
depth
2D retinal image to 3D layout of space
53
monocular depth cues - pictorial
make assumptions about the world
54
Occlusion
Nearer objects block further objects but can’t tell distance apart
55
Relative size and height
Objects that are larger and lower in picture are closer to/ in front
Objects that are smaller and higher in picture are further away
56
Texture gradient
Lots of objects the same size such as bricks or blades of grass gives a stronger sense of depth
57
Familiar size
Use the known size of an object/ person to infer size and distance of an unknown object
58
Linear perspective
Parallel lines in real life converge in 2D pictures
Ponzo illusion - 2 bars the same size look different when placed above converging lines.
59
Name the 4 pictorial depth cues
1. Occlusion
2. Relative size and height (texture gradient)
3. Familiar size
4. Linear perspective
60
Motion parallax
Closer objects move a bigger distance on retina and further objects move slower in background
61
Vergence and accommodation
As focus adjusts and eyes move
62
Binocular disparity - stereopsis
Our eyes are offset in space
- focus on one thing which lands on the fovea of the eyes
63
Crossed disparity
Closer than fixation points on eye (outside the fovea)
64
Uncrossed disparity
Farther than fixation points on eye (falls inside fovea)
65
Gestalt grouping
1. Proximity
2. Similarity
3. Good continuation
4. Closure
66
Kanizsa triangle
3 Pac-Man line up and we see a triangle even though there’s no triangle
- modal completion from amodal completion.
67
T junctions
Signal occlusion
- amodal completion
68
Generic view principle
Not likely things would line up by accident - must be intentional and so the brain amodally completes the object.
69
6 different classes of opponent cells
R+ and G-
G+ and R-
B+ and Y-
Y+ and B-
White+ and Black -
Black+ and White -
70
Spatial antagonism
On and off cells cancel eachother out in specific wavelengths.
Single colour opponency is found in retina and LGN.
71
Double opponent centre surround
Combines different single opponent inputs such as red-green and blue -yellow
72
Assumptions about shadows
Shadows darken surfaces without changing the colour.
“Checker shadow illusion” shows squares are the same colour just lighter or darker so they look different.
73
Context affects colour perception
We assume the colour or brightness of an object based on what we know (is it in light or in shadow?)
74
Motion
A change in spatial position over time.
75
Kinematogram
Motion can be perceived independent of object recognition
76
Motion can be perceived separate of object recognition
Kinematogram
77
Random dot Kinematogram
Motion perception can come before shape recognition
When the dots are static they do not represent a shape, but when they appear to spin they create a cylindrical shape.
78
Waterfall illusion
Variant of the motion after effect (MAE)
After looking at a waterfall then looking away, things still appear to be moving.
79
The motion after effect (MAE)
After adaptation, perceived motion of stationary pattern in opposite direction occurs.
80
Repulsive after effect
Pushes perception in the opposite direction.
The shrinking Buddha example.
81
Reichardt detector
Neuron has a delay so that the motion of an object is perceived at the same time by both eyes.
- if the object crosses the path of one eye first, then there is a delay in time after that position in space until it crosses the path of the other eye so they reach the motion detector unit at the same time.
82
Motion detectors in V1
Combined simple cell receptive fields
- offset in space and time
83
What can the visual system not distinguish between?
Continuous motion and a discrete jump
84
Apparent motion
Perceived smooth motion from alternating stationary targets.
85
Lilac chaser experiment
Green dot is a negative after image.
Can change how smoothly the dots appear to move with different speeds.
86
The correspondence problem
Example: bistable quartet
- unsure whether the dots are moving side to side or up and down based on the distance apart from each other.
87
Local motion
Ambiguous
V1 neurons are tuned to different directions.
88
The aperture problem
Viewing through a small peephole makes it more difficult to see which direction the motion is actually travelling.
89
MT area (V5)
MoTion
Middle Temporal area
Solves the aperture problem because it integrates global motion
90
Global motion
Larger receptive fields
Independent of local direction
Opponent coding of motion direction
91
Motion blindness
Akinetopsia
92
Akinetopsia
Motion blindness
93
Patient LM example of motion blindness
Had a stroke
Damage to MT
Cannot detect motion
Can see water in a glass but cannot see it filling up so it overflows
Her world is made up of disjointed still images - like a slow camera flashing every 3 seconds.
94
Motion perception in daily life
Real world consequences of low-level motion perception
Carries info about objects:
- form
- depth
- biological entities
95
The “deer in headlights” look
Freeze response in many animals
Part of a camouflage strategy
- camouflage is broken by motion
- avoiding predators and approaching prey
Freezing response reinforces the camouflage effect which is conserved across many mamal species.
96
Depth from motion
Motion information gives us clues about form.
-Stereokinesis = strong sense of 3D object when it moves.
-Kinetic depth effect = looks 2D until you rotate it.
-Motion parallax
97
Stereokinesis
Strong sense of 3D object when it moves
98
Kinetic depth effect
Looks 2D until you rotate it
99
Form from motion example
“Point - light displays”
- strong sense of form even from sparce display
Appears to be a bunch of dots until it moves, then it represents a dog playing or a human walking etc…
100
Motion from action
Motion signals arise from interactions with environment
101
Environmental motion signals
“Optical flow”
- as we move around, we gain useful information about how to interact with the environment.
102
Optic flow and heading
Pattern of motion in visual field
Information about direction of motion, distances, direction of movement (heading).
103
Focus of expansion (FOE)
Central point where there’s no movement - tells you direction of heading
104
Heading
Focus of expansion is where we are looking.
Direction of movement is where the body is facing.
- head direction is forward but heading slightly right
105
MT versus MST
MT receptive fields cover only one hemisfield at most.
MST has very large receptive fields which covers both hemisfields.
106
Complex motion
MST neurons respond to complex motion:
- expansion
- contraction
- rotation
107
Uses of optic flow
If you’re not moving, there’s no optic flow pattern.
- self motion cue
- posture control
- time to collision
108
Self motion cue
Detect which way you’re moving
109
Time to collision
Detect objects moving towards you
110
Vection illusions
A lack of optic flow
- optic flow in periphery overrides vestibular input - dominance of vision over vestibular information.
111
Vestibular input
Any motion, movement, tilt or change in direction of the head.
112
Visual control of posture
Vestibular system helps with balance
113
“Swinging room” experiment
- posture
Stationary floor but moving walls and ceiling
- when room moves forwards, gives visual perception that you’re falling backwards
- adjust posture to lean forwards which makes you fall.
114
Time to collision
Object approaching observer - how much time before contact?
- optic flow works out the time to collision and symmetrical expansion predicts a direct hit
115
Time to collision - patterns
When an object is coming towards you, if the visual system is looming or expanding, then it isn’t coming directly at you - if it is symmetrical as it expands then it is coming directly at you.
116
Maintaining collision path
“Linear optical tracking”
- move so that the ball looks like its moving in a straight line, then you should end up in its path.
- fixed line of sight
- symmetrical looming
117
Predicting the present - moving objects
Visual system predicts where the moving object will be.
- doesn’t work for flashing objects
- “flash-lag effect”
118
Assumptions
Brain makes assumptions abut the world:
- paint vs shadows
- depth
This is the visual system’s best guess of what’s out there based on past visual experiences.
119
2D retinal images become 3D
Cues to depth are learnt from experience
120
Monocular depth cues
Monocular = one eye
121
Binocular depth cues
Binocular = two eyes
122
Monocular depth cues examples
- pictorial depth cues
- motion parallax
- accommodation and convergence
123
Accommodation and convergence
How our eyes focus and the movement of the eyes to adjust the lens
124
Pictorial depth cues
-occlusion
- relative height and size
- linear perspective
- aerial perspective (haze)
125
Occlusion
Closer objects block objects further away, but don’t know the distance between these objects from that.
126
Relative size and height
Closer objects appear bigger and lower in picture.
Further objects appear smaller and higher.
127
Familiar size
Use size of known object to infer size of unknown object.
128
Linear perspective
Vanishing point of two parallel lines converging in 2D.
129
Aerial perspective/ “haze”
Objects in distance appear hazy because there’s a lo of air particles between our eyes and that object.
130
Motion parallax
Requires head motion
- closer objects move across field quicker
- further objects move across field slower
131
Vergence
Both eyes move inwards or outwards (non-conjugate movement)
132
Convergence
Near focused
- moves from far away to close
- pulls eyes inwards
133
Divergence
Far focus
- moves from close to far away
- pulls eyes outwards
134
Accommodation
Thickness of lens adjusted depending on focus
- near focus = fatter lens
135
Size illusions
- ponzo illusion
- moon illusion
- Ames rom and Beuchet chair
136
Ponzo illusion
Bars are the same size but placed over converging parallel lines so they appear different sizes.
137
Moon illusion
Relative size and distance cues on horizon
- makes the moon appear larger than usual
138
Ames room illusion
Look through one peephole into a room which looks normal .
- room actually a trapezoid but looks normal from jut one angle.
- makes people of the same height (identical twins) look extremely different heights.
139
Beuchet chair
Chair is in 2 parts but from the angle of the peephole it looks like one whole chair.
- makes people seem insanely different sizes when really they are just certain distances apart.
140
Stereoscopic vision
Depth information from binocular disparity.
141
Binocular vision geometry
Object being fixated on falls on the fovea in both eyes.
142
Horopter
A locus of all the points in 3D space that fall on corresponding retinal points.
143
Binocular disparity
Objects that aren’t on the horopter fall on different retinal points.
144
Crossed disparity
Object is closer so falls on the outside of the fovea in each eye when fixated on.
145
Uncrossed disparity
Object is further away so sits on inside of fovea in each eye when fixated on.
146
Binocular receptive fields in V1
Show sensitivity to both position and phase
- many V1 neurons receive information input from both eyes.
147
Disparity - selective neurons
Different neurons are tuned to different disparities
- receptive fields in each eye
- crossed or uncrossed
148
Stereoblindness
Amblyopia (lazy eye)
- one of the eye’s ocular motor muscles don’t work so the eyes don’t track together.
- only one eye is fixating on what you want it to.
- causes blurred image in visual system.
149
“Stereo Sue” example of stereoblindness
Learnt to see in 3D at age 48
- best treated before the age of 5 if born with it
- wear a patch on good eye to train the bad eye to take input into the visual system.
150
Panum’s fusional area
Region where stereoscopic depth is perceived.
- range of disparities over which fusion of views occurs
151
Outside Panum’s fusional area
Suppression
Double vision
Binocular rivalry
We can’t see our noses because the images don’t match an fall outside the fusional area.
152
Binocular rivalry
When the input to two eyes is completely different
- visual system cycles back and forth between the two images because they’re so different.
- will see a face then a house then a face then a house instead of a face house merged.
153
Seeing stereo
Free fusion
Stereoscope
Anaglyphic glasses
154
Free fusion
Shift focus of eyes to see different images
Parallel or cross-eyed
155
Parallel free fusion
Wide-eyed: focus behind the actual image
156
Cross-eyed free fusion
Focus in front of the image
157
Anaglyphic 3D
Range of disparity
Larger distances produce exaggerated effects
- cut out effect
- miniaturisation
158
Mid-level vision
Image
Surfaces
Objects
159
Image
Orientation
Spatial frequency
Colour
Disparity
Motion
160
Surfaces
Grouping
Occlusion
Completion
161
Surface perception
Surfaces are actively constructed by the brain - going beyond image data.
162
Gestalt principles
Principles of grouping:
- proximity
- similarity
- good continuation
- closure
“Laws” reflect the probability that features go together
163
Proximity
Features near each other are more likely to group together than those that are separated.
164
Similarity
Features that are similar to each other are more likely to group together.
165
Good continuation
Features forming the smoothest contour group together
- smooth contours appear more in the real world
- imagine following a string of
166
Closure
Features that enclose space group more strongly than those that do not.
- dashes create a circle even though that circle’s outline isn’t fully complete.
167
Consequences of closure
Surface completion
- modal and amodal completion
Tendency to complete figures even when not given all the information.
- a panda could have wings but we don’t see the panda with wings we just imagine it based off our experience of pandas.
168
Modal completion
Surfaces appear to continue infront of other objects even though the border isn’t visible.
- relatively uncommon in real life.
169
Amodal completion
Surfaces complete behind occluding surfaces
- not seen but registered
- we know people have faces but if they’re covered we do not know what their face looks like.
It is the visual systems best guess.
170
Amodal completion at work - border occlusion
Allows fragments to be grouped when they otherwise wouldn’t be:
- a group of shapes aren’t connected until border occlusion occurs, ten we group the fragments together to see the shape they make behind the border.
171
Local rules, not object knowledge
Completion is primitive and automatic
- see a naked man behind the dots rather than imagining him in a speedo
- all black squares are equal until you put a red to over some, then some appear to make a larger black square or a black cross.
- 2 donkeys next to each other look normal but cover the middle and it looks like one long donkey
172
Modal from amodal completion
Amodal completion leads to modal completion
- kanizsa triangle
- see the edges of the triangle which makes it look brighter even though here’s no triangle there at all.
173
Edge labelling
Focus on edges of simple block lines and corners
174
Depth edge
Occluding edge
- whatever object is to the right of the arrow (>) is in the front.
- symbolised with > on each edge showing the direction.
175
Convex edge
Edges that stick out towards you
- symbolised by + on the edge that sticks out
176
Concave edge
Edges that retract and point away from you
- symbolised by - on the edge pointing away
177
L vertices
Where 2 lines meet to make an L shape
178
Arrow vertices
Where 2 surfaces meet (3 lines total forming an arrow)
- corner points away from me
179
T vertices
Three lines meet to form a T shape (can be sideways)
- used to tell depth
- two of the lines have to be co-linear
180
Y vertices
Three lines meet to bring together three surfaces
- corner points towards me
- lines make some similarity to a Y
181
T junctions and modal completion
L-junction = an edge
T-junction = occlusion
182
Border ownership
Borders between objects must belong to only one of them objects
- the object that owns the border must be the object in front.
183
Unbounded regions
Regions that do not own the border - usually behind the object which owns the border.
- 2 unbounded surfaces can join together through amodal completion (our visual system may connect them even if they’re not connected)
184
Area V2
Illusory contours
Border ownership
185
Illusory contours
V2 neurons respond to orientation defined by real or illusory contour
186
Neural coding of border ownership
V1 simple cell codes for light and dark but border ownership cell reacts when it’s in the border side rather than the light side.
187
Pattern seeking
The brain is a pattern seeker
- looks for patterns in visual input
- certainty over ambiguity
188
Building surfaces from images
“Quick and dirty” rules - our visual system invents structure from images.
- sometimes there’s more than one interpretation which the brain switches between.
189
Ambiguous figures
Figure-ground reversal
Ambiguous depth
Conceptual ambiguity
190
Figure-ground assignment
Figure = object in the front which owns the border
Ground = surfaces/ objects behind
- sometimes its ambiguous and you can switch between what’s in front and what’s behind.
191
Figure-ground reversal
The face-vase example or silhouettes and columns
- can switch between both but cannot see both together - one always falls into the background.
192
Ambiguous depth perception
Schroeder staircase - can be perceived from the top of stars or below staircase.
193
Conceptually ambiguous figures
Edge labelling is the same but interpretation differs:
. Duck-rabbit
. Young-old woman
. Old man turns into two people getting married
. Mans face turns into two people kissing
Can switch between these images even when you know both.
194
Cues to surface perception
“Quick and dirty” rules
- surfaces don’t line up by accident
- unconscious and automatic
Based on local information, not object knowledge
195
Light and shading
Assume that light comes from above
- dimples and dots are seen due to shading in opposite directions.
196
No accidental alignments
Two edges are unlikely to be accidentally aligned so visual system assumes that they are actually touching.
- man kicking leaning tower of pisa
197
Impossible objects
Penrose triangle
- geometry isn’t physically possible but visual system perceives a 3D object
- even after we know it isn’t connected we still see it as being one shape
- can be edge labelled without problem
198
Impossible objects example
Devils fork
- brain uses knowledge of the world and reality to make them look normal even though they cannot physically exist.
199
Generic viewpoint
Assumes scene is viewed from a generic rather than an accidental viewpoint.
