Form, Depth, Colour, Motion Perception.pdf

Perceiving motion and events Dr Luke Jones Weblinks ◼ http://www.liv.ac.uk/~marcob/Trieste/aperture.html ◼ http://elvers.us/perception/aperture/ ◼ Induced movement ◼ http://psychlab1.hanover.edu/Classes/Sensation/induced/ ◼ Motion after - effects ◼ http://www.michaelbach.de/ot/mot - adapt/index.htm...

Perceiving motion and events Dr Luke Jones Weblinks ◼ http://www.liv.ac.uk/~marcob/Trieste/aperture.html ◼ http://elvers.us/perception/aperture/ ◼ Induced movement ◼ http://psychlab1.hanover.edu/Classes/Sensation/induced/ ◼ Motion after - effects ◼ http://www.michaelbach.de/ot/mot - adapt/index.html ◼ https://www.youtube.com/watch?v=Bholar6Tuhg ◼ Walker ◼ http://www.biomotionlab.ca/Demos/BMLwalker.html ◼ http://www.biomotionlab.ca/Demos/scrambled.html ◼ http://brain.mada.org.il/Shape/shape.html Globe 2Aperature problem 3Evolutionary Importance ◼ Probably evolved very early ◼ Movement = life ◼ Predators that can detect movement of prey more likely to catch it ◼ Prey that can detect movement of predators more likely to survive ◼ Many animals have very poor depth, shape, colour perception ◼ NONE lack the ability to perceive movement (Gordon Wallis, 1942) 4Functions of motion Movements attracts our attention (wave) (active or passive) Movement of an object relative to an observer provides information about object’s 3 D shape. Movement provides information that helps segregate figure from ground and perceptual organisation (common fate) Movement breaks camouflage (freeze reflex) Movement provides information that enables us to actively interact with environment. Ball games Informs of your heading and time to collision, your movement as well as other objects 5Motion and form perception Do we need to be able to recognise an object in order to see it move? Do we match edges and contours between successive views of an object? Is this how motion works? 6Random dot kinematograms suggest not - motion analogs of Random dot stereograms. Instead of presented each simultaneously to the right and left eye, we now present the first and then the second after a short time lag. Here we perceive a central square to move rightward, even though we cannot perceive a square in either frame aloneMotion and form perception 7 8 9 10The 'correspondence problem' highlighted by RDKs suggest that motion detection is direct. We cannot imagine a visual system matching point for point over time in these displays.Motion and form perception 11(1) Real movement (2) Apparent movement (3) Induced movement (4) Autokinetic movement (5) Movement aftereffects 5 Ways to make a spot of light move 12(1) Real movement Light physically moves, i.e. is physically displaced from one place to another. A B 13(1) Real movement - We perceive movement when the eyes are stationary, so that the image moves across the retina. - When an image moves across the retina, it stimulates a series of receptors. - There are neurons in visual system that respond best when a stimulus moves in a particular direction. 14Movement detectors When an image moves across the retina, it stimulates a series of receptors. Excitation and inhibition interact to create a cell that responds only to movement from right to left 15 16 17 18 19 20 21 22 23◼ Detectors such as these have been found in insects and frogs ◼ We have something similar ◼ Cells in cortex sensitive to different orientations, speed and direction of movement ◼ Aperture problem means output of all detectors must be integrated at some stage (Medial Temporal Area ◼ DEMO ch 8 24◼ As the correlated movement of random dots increased, neurons in MT fired more. MT (Middle temporal area) ch 8 25◼ Let the monkey learn to point to the direction of the movement. The monkey pressed the “Up” button • Stimulate the monkey’s MT neurons without showing the visual stimuli. The monkey pressed the “Up” button 26(1) Real movement Much work focused on determining factors that influence: - Threshold for perceiving movement - Perception of velocity Threshold For Detecting Movement 27 28Threshold for movement detection Depends on object and its surroundings, e.g. with the dot and surroundings (e.g. add vertical lines in space between A & B, lower threshold) A B 29 Threshold for detection would be 1/6th to 1/3rd of a degree of visual angle per second (14 seconds) 3 cm of travel, viewed from 30cm 30 Threshold for detection would be as low as 1/60th of a degree of visual angle per second (280 seconds) Perception of Velocity 31 32- perception of motion velocity . Affected by surroundings plus size of both the moving object and framework through which it moves. e.g., a cat in a large cage must move faster than a mouse in a small cage – if they are to appear to move at the same speed. (Helmholtz) Perception of Velocity 33 Circle of left has to travel twice as fast to appear at same speed a circle on the right 34 Circle of left has to travel twice as fast to appear at same speed a circle on the right 35 Cat has to travel twice as fast to appear at same speed as mouse 36 Cat has to travel twice as fast to appear at same speed as mouse 37Movement detectors Cannot explain movement perception when: (1) There is no movement on the retina – as when you follow a moving object with your eyes so your eye movements keep the object ’ s image stationary on fovea. (2) When you perceive no movement when there is movement on retina - as when you move your eyes to look at different parts of the scene or as you walk through a scene. 38Motion perception Need a mechanism that tells us whether retinal stimulation results from movement of stimulus, movement of observer or both: Helmholtz ’ s outflow theory (von Holst, 1954 – Corollary discharge theory) 39Helmholtz ’ s Outflow theory 40◼ If there is a difference between muscle movement command and movement of image across the retina then we perceive movement ◼ E.g. when tracking car, eyes move but retinal signal remains stationary, therefore perceive movement of the car ◼ When keeping eyes still and object moves across we perceive movement ◼ When we look around the world, eye movement command and retinal image movement are equal so we perceive NO movement 41Helmholtz outflow theory Convincing evidence from: 1. Afterimages move when we move our eyes (Eye muscle movement signal no retinal movement). 2. The world moves when we passively wobble our eyes (retinal movement, no eye muscle movement signal). 3. Immobilizing eye - ball results in attempted eye - movement leading to apparent movement of world in opposite direction (Eye movement signal, no retinal movement). ES = Eye movement signal RM = Retinal movement 42(stroboscopic movement) (2) Apparent movement Illusion of movement between two lights by flashing one light on and off, waiting between 40 & 200 msec, then flashing other light on and off. Perception of movement in film = series of static images. 43 44 45 46(2) Apparent movement Less than 30 msec = no movement, simultaneous. Above 30 – 60 msec = partial movement About 60 msec = optimum movement About 60 – 200 msec = Beta and phi movement (phenomenon). Beta: While movement appears to occur between the two lights, it is difficult to actually perceive an object moving across the space between them. Phi perceive an object between Above about 200 msec = no movement, successive Graham (1965) 47 48 49 50 51 52 Slow apparent motion can be ambiguous 53 54 55 56(2) Apparent movement cont... Distance between two lights also affects perception of apparent movement As distance increases, either the time interval or the intensity of the flashes must be increased to maintain the same perception of movement. 57(3) Induced movement Surround spot with another object and then move this object. Duncker (1929) 58 59 60 61 62 63 64 65 66 67 68 69 70 http://psychlab1.hanover.edu/Classes/Sensation/induced/ 71(3) Induced movement Sitting on train – feel it move backward, only to realize that your train is actually standing still, and the one next to you is moving forward. . 72(4) Autokinetic movement Turn out all room lights. When the surrounding framework of the room is not visible, the small stationary light appears to move, usually in an erratic path. 73(4) Autokinetic movement - Sherif (1935) Individually – dot moved from 0.8 – 7.4 inches Group – all reported moved 4 inches Autokinetic effect open to suggestion! - Control of eye muscle not completely stable in dark? Knowing where eyes are and how stable they are may be difficult in dark. - Foo - fighters 74(5) Movement after - effects If an observer first views a pattern moving in one direction, and then views the spot of light, the spot (and surroundings) will appear to move the opposite direction. 75(5) Movement aftereffects Waterfall illusion Anstis & Gregory (1964) – depends on movement of stripes across retina. Supports idea of movement detectors, which respond only to movement across the retina. GH 76 77http://www.lifesci.sussex.ac.uk/home/George_Mather/Motion/MAE.HTML http://www.michaelbach.de/ot/mot_adapt/ http://www.psychologie.tu - dresden.de/i1/kaw/diverses%20Material/www.illusionworks.com/html/motion_aftereffect.html http://dogfeathers.com/java/spirals.html Motion after - effect demos http://www.youtube.com/watch?v=Vn1GaaLhz4g&feature=related 78Hubel & Wiesel (1959) identified directionally specific motion detectors. Sutherland (1961) argued that motion after - effects arose from an imbalance in the ratio of activities from two sets of directionally - tuned receptors, each sensitive to the opposite directions of motion. Barlow & Hill (1963) provide direct evidence to support ratio hypothesis. Ratio hypothesis 79Barlow & Hill (1963) Impulses per second ON OFF TimeFiring in preferred direction cell Firing in null direction cell Period of presumed motion after - effectNote fall off in rate of firing 80Event perception Movement provides information about 3 D shape, helps us segregate figure from ground, and interact with the environment. Examples: - Creating structure from motion 81Creating structure through motion 82http://www.biomotionlab.ca/Demos/BMLwalker.html http://www.biomotionlab.ca/Demos/scra mbled.html 83 84Not just biological motion…. http://brain.mada.org.il/Shape/shape.html 85Motion Induced Blindness 86 http://lite.bu.edu/vision/applets/Motion/Bl indness/Blindness.swf 87 Colour vision 2Colour vision (1) What is colour good for? (2) Light and Colour (3) Theories of Colour Perception (4) Colour Blindness 3Of all mammals, only a few species have the necessary hardware to see “ colours ” in a way comparable to us (trichromatic). • Catarrhine monkeys (Old World monkeys and man) • Platyrrhine monkeys (New World monkeys - only the females) • Some tropical fish and birds have 4 types (tetrachromacy ) • Pigeons have 5! (pentachromats) (1) What is colour good for? 4(1) What is colour good for? (1) Scene Segmentation: Variations in colour often signal object boundaries (2) Camouflage: Animals use this fact to disguise themselves by colour markings (3) Perceptual Organisation: Our visual system uses colour to group elements in a scene 5• Also strong evolutionary force depending on species • Food identification – Ripe fruit – Correct leaves – Harmless or harmful berries – Poisonous or venomous animals 6Camouflage 7Camouflage 8• Visible light forms a narrow band of frequencies in the electromagnetic spectrum. • Within this band, different frequencies (or wavelengths) have different hues, ranging from red (for long wavelength light) to violet (for short wave length light). What is colour? 9Electromagnetic Spectrum 1 millimeter = 1 000 000 nanometers Whole spectrum of visible colours covers just 400 nm 10• 0.00000000000001 meters to 10,000 meters • We see between 0.00000390 and 0.00000750 meters • range of 4 thousandths of a millimeter 11 12 13• Different objects absorb and reflect different wavelengths of light • This gives them their ‘colour’ • The colour also depends on the light source 14• The phenomenon of colour is more complicated than just wavelength judgement. • The wavelength of the light reflected only determines the hue which is seen. • Perceived colour is also determined by the: – intensity of the reflected light (how bright it is) – the saturation of the colour (how much white light is mixed in with the pure hue). Hue, intensity and saturation 15(2) Light and Colour Property of Light Psychological Attribute Wavelength Hue (colour) Intensity Brightness (perceived intensity) Spectral Purity Saturation (i.e how much colour or how much white) 16(2) Examples 17 18( 3 ) Theories of Colour Perception Trichromatic Theory Opponent Process Theory 19Trichromatic Theory (Young - Helmholtz) There are three receptor types and their combined responses account for all colours Blue - sensitive cones maximally responsive to short wavelengths. (S - Cones) Green - sensitive cones maximally responsive to medium wavelengths. (M - Cones) Red - sensitive cones maximally responsive to long wavelengths. (L - Cones) 20Not exactly red green & blue! So we usually now refer to them as S, M & L 21Short: ~ 419 nm Medium: ~ 531 nm Long: ~ 558 nmThree types of Cones: S, M and L S M L 22 23 24Tapetum lucidum (tap - e - tum lucyd - um) 25 26Pigments in the cones determine which wavelengths they respond to best 27Hang on a minute... • From the cones light is encoded as electrical signal • Light is not coloured, objects are not coloured and neither is the visual signal Is your Red the same as my Red? • Explanatory gap • Example of pain • Consider tastes in colour • Consider differences in taste. 28 29Retina: cone distribution ◼ There is a concentration of cones in an area called the fovea – a small pit in the retina (edge to edge: retina 32mm, fovea 1.5mm). ◼ Eye movements – image of object of interest falls on the fovea. ◼ Fovea has highest density of receptors - best acuity (1% - 50%) ◼ No rods in the centre of the fovea. 30Rods versus Cones • Rods very sensitive don’t need much intensity to be activated • Cones need good lighting conditions to be activated • Rods sensitive to intensity (black - white) only • Cones responsive to different wavelengths - 3 types • Rods found all over retina except centre of fovea • Cones concentrated in fovea • 20 times more rods than cones across whole retina 31Trichromatic Theory (Young - Helmholtz(1802:1866) • This theory hypothesised that there are three different sorts of receptors and that they respond best to different wavelengths of light. • They respond best to long wavelength (which looks yellow/ red), medium (green ) or short (blue) wavelength light - this is what cones do! • The colour you see is determined by the relative levels of activity in the three sorts of receptors. • So red objects reflect more long wavelength light than other wavelengths. 32Support for Trichromatic Theory • Three primary colours combine to produce all possible colours • Three forms of dichromatism (colour blindness) A mixture of green and red light produces same perception of yellow colour as monochromatic yellow light (metamersim) - we can t tell the difference 33Afterimages 34Only red and green cones can respond producing Yellow. Blue cones/channel fatigued/adapted 35Opponent Process Theory (Hering (1920), Hurvich - Jameson) Hering noticed that when people are presented with large number of colour samples and asked to pick out those that are pure (not a mix), then: The pick a red a green and a blue (as predicted by trichromatic theory) But also Yellow! Also cones and fatigue were not understood, so it was unclear how trichromacy could explain afterimages 36So he proposed 37Opponent Process Theory (Hering, Hurvich - Jameson) 3 processes which are opponent in nature: (i) Red - Green (ii) Yellow - Blue (iii) Black - White e.g. Red - Green Receptor will signal either Red or Green but not both 38 39Short: ~ 419 nm Medium: ~ 531 nm Long: ~ 558 nmThree types of Cones: S, M and L S M L 40Support for Opponent Process Theory (1) Non - existence of certain colours, e.g., bluish - yellow (2) Colour confusions in colour blindness (red and green) (3) Complementary afterimages (4) Colour context effects 41Colour context effect 42 43 4444 4545 Lightness constancy Perceive the squares are different even though reflected amount of light is same Lightness constancy 48 Blue and Black? Gold and White? 49 50 51Both theories are in fact correct (a) trichromacy at the level of the cones (b) opponent processes at the level of LGN ( The Lateral Geniculate Nucleus ) and cortical cells.Trichromacy versus opponent process 52 53Colour Blindness • First described by John Dalton in 1794 • Suffered from a form of it himself • Diagnosed 200 years later! 54(4) Colour Blindness Anopias: Insensitive to L, M or S wavelengths of light. (missing a type of cone) Anomalies: Misalignment of L or M in trichromats. (distribution or deficiency) 55'Colour blindness ’ should really be termed 'colour deficiency'. With the exception of people with cortical colour blindness who see the world in B/W, colour blind people just exhibit colour confusions. Their experience of colour is very different from 'normal' (4) Colour Blindness 56Anopia: Dichromatism: Missing cones (a) Protanopia: L - cone pigment missing: 1.3% M 0.02% F (b) Deuteranopia: M - cone pigment missing: 1.2% M 0.01% F (c) Tritanopia: S - cone pigment missing: 0.001% M 0.003% F 57Protanopes (L missing) see only 2, Deuteranopes (M missing) see only 4 58 59 60Anomilies Anomalous Trichromatism (a) Protoanomaly: L - cone pigment deficiency: 1.3% M 0.02% F (need more ‘ red ’ in ‘ red - green ’ mixture to match ‘ yellow ’ ) (b) Deuteranomaly: M - cone pigment deficiency: 5.0 % M 0.35% F (need more ‘ green ’ in ‘ red - green ’ mixture to match ‘ yellow ’ ) 61Note that colour blindness types supports both colour vision theories. 1)Whole fact of anopia points to 3 cone types. 2)Opponent process theory supported by the fact that people who have trouble with RED also have trouble with GREEN etc 62Human Tetrachromats • Some very rare humans (predominantly female) have 4 pigment cone types • Can detect variations in hue that we normally can not. Other animals • Some birds and butterflies have 5 receptors…. • But….. 63 • What might it look like? • Impossible to know • Explanatory Gap • Higher and lower animals, smell, vision 64 Depth Perception Dr Luke Jones [email protected] Seeing a 3 - D world Image from real world is essentially two -dimensional. Yet our perception is of a three -dimensional world. Generally people are quite accurate in judging ambient distance (up to about twenty feet). This is typically demonstrated by having them survey the scene, close their eyes, and walk to a predesignated object (Loomis et al., 1992). 2 Strong Innate Element to Perception of Depth ? 3 Visual Cliff video • http://www.youtube.com/watch?v=DrzmvI6iMrE • From 53 sec 4 Cues to depth (1 ) Oculomotor cues: cues that depend on our ability to sense the position of our eyes and tension in our eye muscles. (2) Pictorial cues (monocular cues) : cues that can be depicted in a still picture. (3) Motion -produced cues: cues that depend on movement of the observer, or movement of objects in the environment. (4) Binocular disparity: a cue that depends on the fact that slightly different images of a scene are formed on each eye. 5 1) Oculomotor Cues cues that depend on our ability to sense the position of our eyes and tension in our eye muscles (1) Oculomotor cues DEMONSTRATION Look at your finger as you hold it at arm’s length. Then slowly move your finger towards your nose. Be aware of how, as your finger moves closer, you feel your eyes looking inward and you feel increasing tension inside your eyes. 7 (1) Oculomotor cues Feelings you experiences are: - convergence as your eye muscles cause you eyes to look inward - accommodation as the lens bulges to focus on a near object) Shape of lens and position of eyes are correlated with the distance of the object we are observing. Only effective cues at distances closer than 5 – 10 feet (Liebowitz, Shina & Hennessay, 1972). 8 9 The closer the object the greater the convergence 10 11 2) Pictorial Cues Cues that can be depicted in a still (2D) picture. 12 2) Pictorial Cues • Pictorial Cues (also called monocular cues) do not require viewing with both eyes in order to work • In fact often better to view monocularly • TV, photos, paintings 13 i) Overlap or Interposition or Occlusion • One object obscures part of another, or overlaps with it 14NB: Gestault Completion 15 (2) Pictorial cues 16Place des Lices Paul Signac, 1893 (i) Overlapl/Interposition 17Overlap cont.. Place des Lices Paul Signac, 1893 (ii) Relative Size 18 The retinal size of objects gets smaller as they get further away 19Perception of size and depth perception ◼ An object can look the same size at different distances But: retinal image size changes with distance ◼ Increase distance : decrease retinal image size ◼ Decrease distance : increase retinal image size ◼ The fact that an object can look the same size regardless of changing retinal image size is referred to as size constancy SIZE CONSTANCY 20 21 Emmert ’ s law • Emmert's Law states that objects that generate retinal images of the same size will look different in physical size if they appear to be located at different distances. • Specifically, the perceived size of an object increases as its perceived distance from the observer increases. • An object of constant size will project progressively smaller retinal images as its distance from the observer increases. • Similarly, if the retinal images of two different objects at different distances are the same, the physical size of the object that's farther away must be larger than the one that is closer. 22 (iii) Relative Height 24• As objects get further away they get nearer the horizon • IF the objects are below eye height the highest object is furthest away • If the objects are above eye height then the lowest object • is further away 25 26• If the objects are above eye height then the lowest object is further away 27• IF the objects are below eye height the highest object is furthest away (iv) Atmospheric Perspective 28Distant objects appear less sharp because more air and particles to look through. 29 Also appear more blue as blue light is scattered more by atmosphere (iv) Familiar Size 30 31(v) Linear perspective Lines that are parallel in the scene converge as they get further away. But remember the retina is curved! 32 33 Shadows within objects - Attached Shadows The shading that results from depth within an object is a cue to depth i . e . ATTACHED SHADOW . The meaning of shading is ambiguous . A depression and an elevation of a surface will be shaded on one side . Telling the difference relies on knowing the direction of the light source . We assume that light comes from above . 34(vi)Shading and shadow 35Assume direction of lighting from above. 36 37 38 Detached Shadows 39 40 Texture gradient 41 Texture becomes smaller/finer as distance increases 42 Texture gradient cont… 43 (3) Movement - produced cues (i ) Motion parallax As an observer moves relative to a 3 -D scene, nearby objects appear to move rapidly whereas far objects appear to move slowly. Monocular cue to depth. http://psych.hanover.edu/Krantz/MotionParallax/MotionParallax.html 44 45 1. Relative direction 2. Amount of motion 46 47 Motion Parallax • Used more by animals that don’ t have much binocular overlap. • Head bob and orthogonal running 48 49(3) Movement - produced cues (ii) Deletion and Accretion As one object moves in front of another, deletion occurs whereby the front object covers more of the back object. As one object moves away from another, accretion occurs whereby the front object covers less of the back object. 50 2. Accretion & Deletion 51 Deletion Accretion Binocular Disparity 52 (4) Binocular disparity • Also called binocular stereopsis. • Cue depends on two eyes & fact that our eyes see the world from slightly different positions determined by the distance between them. • Basis of stereoscope (Wheatstone, 1802 - 1875) & 3 - D movies. 53 (4) Binocular disparity DEMONSTRATION With only your right eye open hold one finger upright about 6 inches in front of you. Then position a finger from your other hand about 6 inches further back, so that it is completely hidden by the front finger. Now close your right eye and open your left one, and the rear finger become visible. Since your left eye sees from a different point of view, it has looked around your front finger. 54 55 (4) Binocular disparity When two eyes receive slightly different images of the same scene, we experience an impression of depth. Why? Corresponding retinal points…. For every point on one retina, there is a corresponding point on the other. These points would be identical if one retina was moved over to superimpose the other retina 56 57 Corresponding Retinal Points Regions on the two retinae that would overlap if you slid one retina on top of the other. When you fixate on an object it will stimulate corresponding points in the two eyes. 58 Non - Corresponding Retinal Points Regions on the two retinae that would not overlap if you slid one retina on top of the other. These points are separated on the retinae, and create disparity. 59 Fixation Point Corresponding Point at different depth Non - Corresponding 60 61The red lines between the tree and the retinas describe the angle of convergence the eyes make when the tree is imaged on both foveae. The tree is imaged on corresponding points of the retina - there is no disparity. Because the eyes are fixated on the tree, the policeman, has its image falling on non - corresponding points of the retinas. That is to say its images are binocular disparate. (4) Binocular disparity An object located between the policeman and the tree would be result in less disparate images on the retina. The amount of disparity tells us how far the policeman is from the tree. 2 - 5% of people show stereo -blind performance and appear to lack mechanisms for processing disparities (Richards, 1971). 62 Binocular Disparity • This cue for depth diminishes with distance • Determined by distance of the two eyes • Hyperstereo – can give increased depth from disparity 63 Hyperstereo Telestereoscope 64 (4) Binocular disparity Julesz (1964) demonstrated that the visual system can use disparity information directly to generate a percept of depth. 65To create a vivid sense of stereoscopic depth: Present the same image to both eyes But, shift one slightly to the left or right The shifted area will appear to be displaced in depth This 3D image is called a stereogram There are many ways to make a stereogram: 66 Random Dot Stereogram 67 Animated autostereogram 68 69 Cube Colour Filters 70 Orthostereography 71 Shutter Glasses 72 Lenticular Displays/Printing 73 Parallax Barrier Technology 74 75 Virtual Reality 76 , 96 Cutting, 1996 Depth Contrast Depth (meters) Occlusion 1 10 100 Size constancy Cast Shadows Disparity Motion parallax Convergence Aerial Perceiving Form and Perceptual [email protected] Perception of Form & OrganisationAims:-Te l l y o u a b o u t w h y t h e p e r c e p t i o n o f f o r m a n d o r g a n i s a t i o n i s important.-Marrʼs approach-Gestalt approach to organisation including laws (for example, Law of Prägnanz) & figure-ground decisions … Perception of Form & Organisation• Why is the perception of form and organisation important?-Environment contains hundreds of overlapping objects.-Ye t p e r c e p t u a l e x p e r i e n c e i s o f s t r u c t u r e d , c o h e r e n t o b j e c t s which we can recognise, use and usually name.-Starting point is that light, reflected from objects, reaches the eyes. How do we go from this perceiving a coherent, stable, 3D object? -Fallacy of the visual system as camera analogy Even this is misleading And remember the retina is curved! And its constantly moving, and being updated 50 times a second! So…•Receptors are unevenly distributed•Image is inverted, distorted and tiny and FLAT•Compensate for eye movement, body movement and object movement•Uneven amount of cortex devoted to different parts of the visual field•Nothing visual about the cortical representation? Nothing square about a square. Men In Black (1997) Perception of Form & OrganisationTw o a p p r o a c h e s :(1)Marr’s approach, concerned with the representation of edges, contours and other areas of contrast change(2) Gestalt approach, concerned with rules of perceptual organisation (1) Marrʼs approach-David Marr (English) 1945-1980 (published ʻVisionʼ1982)-ʻBottom upʼapproach. -Starts with input to perceptual system in form of retinal image and describes the stages in processing of this image.-Each stage takes as its input the information from the previous stage and transforms it into a more complex description or representation.-Computational model Computational Model•Computational theory•What is the model trying to do? What are the processes for? What is the goal?•Algorithmic level•What algorithm is needed? What process?•Mechanism Level•What mechanism is needed to implement the algorithm? I.E. Neural/biological system Rubikʼs Cube Example•Scrambled Cube•Cross Algorithm•Cross Complete On One Side•Corners Algorithm•One side/layer complete•Middle layer corners Algorithm•Middle Layer complete•Cross Algorithm•Cross Complete on final side/layer•Rotate Cross To correct position Algorithm•Cross in correct position•Move corners to Correct position Algorithm•Corners In correct position•Rotate Corners to correct Position•Solved Cube Retinal image is analysed sequentially at different levels:Retinal ImageGrey level description–measuring intensity of light at each point in image.Primal sketch-representation of contrast change (blobs, edges, bars etc) over range of spatial frequencies)21/2D sketch-representation of orientation, depth, colour relative to the observer3D representation-representation of objects independent of observer(1) Marrʼs approach Retinal Input to Grey Level To Primal Sketch Primal Sketch to 21/2D Sketch•Primal Sketch combined with depth cues, colour, motion.•Its not 3D because it is observer-orientated. (unseen parts of scene and objects) 3D representation•2 1/2 sketch analyzed for 3D volume primitives (cylinders, cones, cubes etc).•Produces 3D representation that is independent of observer•Conscious experience of vision. Identify edges and primitivesGroup primitives and processPerceive 3-D objectObject’s image on retinaPrimal Sketch2 1/2D Sketch3D Representation Importance of the Computational Approach•An algorithm/rule/system is more likely to be understood by understanding the problem that has to be solved, rather than the examining the mechanism (and hardware) in which it is embodied•To u n d e r s t a n d p e r c e p t i o n ( p u r e l y ) b y s t u d y i n g n e u r o n s i s l i k e t r y i n g to understand bird flight by studying only feathers : function not form (AI argument) •Gestalt psychology –the whole is greater (different) than the sum of its parts (Max Wertheimer, 1912). •To p-Down Approach•Gestalt psychologists interested in how we group parts of a stimulus together and the way we separate figure from ground….SEGREGATION and GROUPING(2) Gestalt approach Necker CubeMultistability The Gestalt SchoolMax WertheimerKurt Koffka1886-19471880-1943 Wo l f g a n g K o h l e r1887-1967Series of Experiments on Kohler and Koffka by Wertheimer, together they developed the Gestalt schoolSeries of influential publications in 1920’s •Gestalt psychology –the whole is greater (different) than the sum of its parts (Max Wertheimer, 1912). •Don’t see lines and figures but forms and shapes (Gestalt means form/shape in German)•To p-Down Approach•Gestalt psychologists interested in how we group parts of a stimulus together and the way we separate figure from ground….SEGREGATION and GROUPING(2) Gestalt approach Perceptual organisationAmbiguity generally does not arise in the real world. Rather, we usually see a stable and organised world.For example, most people see a set of overlapping circles, rather than one circle touching two adjoining shapes that have ‘bites’taken out of them. Why? •Argue that we see objects according to all their elements taken together as a whole•Sought to isolate principles of perception•Seemingly innate ʻlawsʼwhich determine way in which objectsare perceived Gestalt laws of perceptualorganisation1.Similarity2.Good continuation3.Proximity4.Connectedness5.Closure6.Common Fate7.Familiarity8.Invariance9.Prägnanz –“good figure” 1. SimilaritySimilar things appear to be grouped togetherGrouping can occur due to shape, lightness, hue, orientation, size …. 2. Good continuationPoints that, when connected, result in straight or smoothly curving lines, are seen as belonging together, and the lines tend to be seen in such a way as to follow the smoothest path. Reification -More spatial info than is present 3. ProximityThings that are near to one another appear to be grouped together 4. ConnectednessThings that are physically connected are perceived as a unit (Rock & Palmer, 1990, December, Scientific American). 5. ClosureOf several geometrically possible perceptual organisations, a closed figure will be preferred to an open figure. We t e n d t o ‘ c o m p l e t e ’ a b r o k e n f i g u r e b e c a u s e o f t h e Strong closure cue for organizing what we see Reification -More spatial info than is present 6. Common FateThings that are moving in the same direction are grouped together Common Fate cont..•Objects with same orientation are grouped together 7. FamiliarityThings are more likely to form groups if the groups appear familiar or meaningful.Explanation or description? 8: InvarianceMAJOR Problem inComputer vision CAPTCHA Test•Completely Automated Public Turing test to tell Computer and Humans Apart 8. Prägnanz –“Good Figure”•The central law of Gestalt Psychology•Many of the laws are manifestations of Prägnanz•“Of several geometrically possible organisations that one will occur which possesses the best, simplest and most stable shape”Koffka, K. (1935) Principles of Gestalt Psychology. New York: Harcourt Brace (p138). Figure-Ground SegregationGestalt psychologists also interested in how we separate figure from ground..Usually no doubt –but some reversible figure-ground patterns.Rubin Mosaic II by M.C. Escher(1957)Kitchen example Figure-Ground•These are extreme examples•Normally in a visual scene some objects (figures) seem prominent, and other aspects of field recede into the background (ground).•Lecturer (figure), other objects (background)•Gestalt interested in this because it infers top-down process Figure-Ground Segregation-Properties that affect whether area seen as figure or ground are:-Symmetry: symmetrical areas usually figure.-Convexity: convex shapes usually figure.-Area: stimuli with comparatively smaller area usually figure.-Orientation: vertical and horizontal orientations usually figure.-Meaning/Importance: meaningful objects more likely to be seen as figure. Implies attention -top-down Problems with the Gestalt approach•Underplay the parallel processing and unconscious processing that the brain does•Explanation of how some of their laws worked was wrong.•Their laws provide a description of how things work rather than an explanation.•Their laws are ill defined –Prägnanz –what is the simplest and most stable shape?•Stating the obvious? Some positive points•Their laws actually appear to be generally correct.•Percepts can be analysed into basic elements.•The whole is greater than the sum of its parts.•Context and experience effect perception Bottom-Up versus Top-Down-Bottom-UpStart from the bottom, considering physical stimuli being perceived and then work their way up to higher-order cognitive processes (organizing principles and concepts)-Higher cognitive processes can not directly influence processing at lower levels•E.G. Marr’s computational approach •To p-Down•The perceiver builds (constructs) a cognitive understanding (perception) of a stimulus, using sensory information as the foundation for the structure but alsousing other sources of information to build the perception •During perception we quickly form and test various hypotheses regarding percepts based on•What we sense (sensory data)•What we know (knowledge stored in memory)•What we can infer (using thinking)•What we expect•Example of Orange KEY CONCEPTS•Marr’s approach inc. stages•Gestalt approach•Segregation and grouping•Gestalt laws of perceptual organization•Problems with Gestalt approach•Positive points of Gestalt approach•Figure-ground segregation•Bottom-Up Versus Top-Down