Constructing The Visual World PDF

Constructing the Visual World Psyc 160: Cognitive Psychology Pomona College  Outline Computational problems in vision – Perception as inference and classification Early vision – Retinal anatomy – Image pre-processing Cortical division of la...

Constructing the Visual World Psyc 160: Cognitive Psychology Pomona College  Outline Computational problems in vision – Perception as inference and classification Early vision – Retinal anatomy – Image pre-processing Cortical division of labor – Parts, properties, and locations – Selective deficits  What is perception? EXPERIENC FUNCTI E ON ? Learnin Model of the g Distal Proxima Langua Distal ge Stimulu l Reasoni Stimulus s Stimulu ng s Decisio n … Perceptual systems “infer” and “classify”. Perception is constructive  “Illusions” Patches A and B have the same shading in the image. The 2 monsters are the same size in the image.  The inverse problem path of causation (retinal stimulation) Proximal stimulus Distal stimulus path of inference 2 distal stimuli Problem: (big object far away or short object close up) produce the same proximal stimulation at the retina.  The inverse problem path of causation (retinal stimulation) Proximal stimulus Distal stimulus path of inference 1. Estimate distance of stimulus Strategy: 2. Infer size from distance Monocular depth cues cues available to a single eye* Laws of perspective  Texture density increases into the distance. * When both eyes are used, further information about depth can be gleaned by comparing the two retinal projections (binocular cues) Monocular depth cues Laws of perspective  Parallel lines converge into the distance.  The inverse problem The Ponzo Illusion Visual system “infers” from context that rightmost object is farther away. Since retinal stimulation is identical, this implies that rightmost object is larger.  Perception as “Unconscio us Inference” Hermann Von Helmholtz  “Unconscious Inference” Some “illusions” can be understood as best guesses about the distal stimulus. EXPERIENC E ? Model of the Distal Proxima Distal Stimulu l Stimulus s Stimulu s “what is the likeliest cause of the proximal stimulus?”  Another inverse problem a white surface in dim light reflect the and same intensity of light onto the retina a dark surface in bright light  Another inverse problem local input contextual inference A Dark surface illuminated B Light surface shadowed  A useful way to think about perception. But there is no Perception homunculus making the as inferences! Just neural “Unconscio networks implementin us Inference” g rules to match outputs to Key questions: inputs. The rules often select What computations does vision use? likely causes of ambiguous effects How are they neurally (but no “unconscious insight” is implemented? What happens when 2 distal stimuli are equally likely? Necker cube The perceptual system alternates between equally likely interpretations.  Binocular Rivalry reported percept: “BISTABLE PERCEPTIO time  Inference Problems in Perception A key task of perceptual systems is to “infer” the distal stimulus in the world from the proximal stimulation at the sensory receptor. But there is an “inverse problem”: Many different distal stimuli can produce the same pattern of proximal stimulation. Neural networks implement “heuristics” (rules of thumb), which tend to guess the likeliest distal cause of ambiguous proximal stimulation. This helps explain some perceptual illusions. – Unusual distal causes of proximal stimulation are misidentified as typical causes. – Perceptual systems are “encapsulated”: Non-perceptual knowledge often fails to influence perception. Perceptual systems are largely encapsulated. (Your knowledge of how an illusion works does not eliminate the illusion.) But… In limited ways, non-perceptual beliefs can influence what we perceive.  What is perception? Key questions: What computations does vision use? EXPERIENC FUNCTI How are they neurally E implemented? ON ? Learnin Model of the g Distal Proxima Langua Distal ge Stimulu l Reasoni Stimulus s Stimulu ng s Decisio n … Perceptual systems “infer” and “classify”. Perception is constructive  The Problem pattern of stimulation intensity on retina low-light vision bright-light vision What is Where?  The Problem pattern of stimulation intensity on retina “The first great revelation was that the problems are difficult. Of course, these days this fact is a commonplace. But in the low-light vision bright-light vision 1960s almost no one realized that machine vision was difficult.” Marr, 1982, Vision “We will examine many theories in the course of this book. All of them have one thing in common: They will all ultimately be proved inadequate.” Palmer, 1999, Vision Science What is Where?  It’s not easy being seen. SUBJECTIVE COMPUTATIONAL DIFFICULTY COMPLEXITY high low complex arithmetic chess object recognition language comprehension low high  Outline Computational problems in vision – Perception as inference and classification Early vision – Retinal anatomy – Image pre-processing Cortical division of labor – Parts, properties, and locations – Selective deficits Akinetopsia, Achromatopsia, Balint’s  1. Light 2. Eye 3. Brain visual cornea cortex and lens Retina electromagnetic radiation Different has particle and wave properties cortical areas process visible light: different small part of spectrum aspects of the visual world physical wavelength maps other photoreceptor onto perceived color neurons cells  Layers of the Retina light brainward bound ganglion cells (~1 million) neural signals photoreceptors (> 100 million) back of retina  Photoreceptors Cones Rods Specialized for Specialized for high-light color vision low-light vision Three cone types… A single rod type …with maximal Highly sensitive sensitivity to different wavelengths S M L The visual system Activity “infers” wavelength from the relative activity of the different cone types. wavelength  Rods and Cones cones rods Rods and Cones cone rods peripheral retina of monkey Retinal Geography  Retinal Geography light gangli on cel ls bipolar cells photoreceptors The foveal pit Retinal Geography Fixate subtle stimuli in bright light… …but fixate away in dim light. Retinal Geography fovea head of the optic nerve (“blind spot”)  Exercise: Find your blind spot Left Righ eye t eye f f  Layers of the Retina light brainward bound ganglion cells (~1 million) neural signals amacrine cells bipolar cells horizontal cells photoreceptors (> 100 million) back of retina  Receptive Fields of Retinal Ganglion Cells RGCs respond to light in a small region of the retina. Light has opposite effects in the center v. surround of the receptive field. 2 types: – – + + – + + – – + – + + – – + – + – + – ON-CENTER – + + – OFF-CENTER – + + + + – – – – + + + – – – – + + – – – + – + – + + Ganglion cells make “relative judgments”: Is there more light in the center or periphery of a small patch of retina?  Retinal Information Processing - ---------------------- - ++++++++++++++++ - - +++++++++++++ +++++++++++++ - - how will on- baseline - - center ganglion - baseline - - - cells respond to - - this image? - - - - back of- retina - - +- -+-+- + - -+- +- -+-+- + - -+- +- -+-+- + - -+- +- -  Retinal Information Processing -+ + + + + + + + + + + + + + + + - ---------------------- - - +++++++++++++ +++++++++++++ - - how will off- baseline - - center ganglion - baseline - - - cells respond to - - this image? - - - - CENTER- back of retina - ---------------------- - - + + + + + + + + + + + + + + + +- SURROUND RESPONSES FACILITATE EDGE - ---------------------- - - ++++++++++++++++ - DETECTION +++++++++++++ +++++++++++++ - - IN how will on- baseline - - center ganglion - - SUBSEQUENT baseline - - PROCESSING cells respond to - - this image? - - - - back of- retina - - +- -+-+- + - -+- +- -+-+- + - -+- +- -+-+- + - -+- +- -  Retinal Information Processing -+ + + + + + + + + + + + + + + + - ---------------------- - - +++++++++++++ +++++++++++++ - - how will off- baseline - - center ganglion - baseline - - - cells respond to - - this image? - - - - back of retina - ---------------------- - partly explains - + + + + + + + + + + + + + + + +- illusion - ---------------------- - ++++++++++++++++ - - +++++++++++++ +++++++++++++ - - how will on- baseline - - center ganglion - baseline - - - cells respond to - - this image? - - - - back of- retina - - +- -+-+- + - -+- +- -+-+- + - -+- +- -+-+- + - -+- +- - Each hemisphere represents the contralateral visual hemifield. The Human Visual System primary visual cortex (V1)  Early Vision The “duplex” retina – Three cone types, especially concentrated in the fovea, specialized for bright-light color vision – One rod type, specialized for night-light vision Retinal information processing – Ganglion cells responsive to local light gradients Primary pathway  Outline Computational problems in vision – Perception as inference and classification Early vision – Retinal anatomy – Image pre-processing Cortical division of labor – Parts, properties, and locations – Selective deficits  Primary Visual Cortex “We tried everything short of standing on our heads to get it to fire… To stimulate, we were using mostly white circular spots and black spots…. After about five hours of Hubel Wiesel struggle, we suddenly had the impression that the glass with the dot was occasionally producing a response, but the response seemed to have little to do with the dot. Eventually we caught on: it was the sharp but faint shadow cast by the edge of the glass as we slid it into the slot that was doing the trick. We soon convinced ourselves that the edge worked only when its shadow was swept across one small part of the retina… Most amazing was the contrast between the machine-gun discharge when the orientation of the stimulus was just right and the utter lack of response if we changed the orientation or simply shined a bright flashlight into the cat’s eyes.” Hubel, 1988  Primary Visual Cortex Hubel Wiesel Many cells respond to bars/edges of specific orientations in specific regions of the visual field (sometimes moving in specific ways) Primary Visual Cortex Retinotopic organization of V1 with cortical magnification of fovea Primary Visual Cortex wavelength selectivity Segregation of color and orientation information Organization of Information Coding in V1 The Human Visual System 3. Brain visual cortex Different cortical areas process different aspects of the visual world  3. Brain visual cortex Different Focal brain damage can lead cortical areas to striking selective deficits process of visual perception. different aspects of the visual world  Cerebral Akinetopsia “She had difficulty, for example, in “When I’m looking at the car first, it pouring tea or coffee into a cup seems far away. But then, when I because the fluid appeared to be want to cross the road, suddenly frozen, like a glacier. In addition, the car is very near.” she could not stop pouring at the right time since she was unable to perceive the movement in the cup - Zihl et al., 1983 when the fluid rose.”  Cerebral Achromatopsia "Mr. I. could hardly bear the changed appearances of people (‘like animated gray statues’) any more than he could bear his own changed appearance in the mirror: he shunned social intercourse and found sexual intercourse impossible. He saw people's flesh, his wife's flesh, his own flesh, as an abhorrent gray; ‘flesh-colored’ now appeared ‘rat-colored’ to him. This was so even when he closed his eyes, for his preternaturally vivid (‘eidetic’) visual imagery was preserved but now without color, and forced on him images, forced him to ‘see’ but see internally with the wrongness of From Sacks & Wasserman, 1987, his achromatopsia. He found foods “The case of the colorblind painter” disgusting in their grayish, dead appearance and had to close his Balint’s Syndrome Bilateral destruction of spatial maps in the dorsal stream  Balint’s Syndrome Spatial disorientation Inability to localize visual objects, estimate visual distances (while auditory and tactile stimuli are localized correctly) Optic ataxia Disorder of visually guided reaching Gaze apraxia Impairment of purposeful eye movements Simultanagnosia “When I see one object, I do not see the other and it(single takesseen meobject timecan to be find where large or small) Optic ataxia Simultanagnosia  Outline Computational problems in vision – Perception as inference and classification Early vision – Retinal anatomy – Image pre-processing Cortical division of labor – Parts, properties, and locations – Selective deficits