PSYCH 3420 Prelim 1

53. What does the CSF say about the differences between the visual system of the infant and the adult?

Answer almost identical to 45 Visual acuity increases with age

15. Concepts to understand: Luminance vs. radiance, hyperopia, myopia, bits, blindsight, blind spot, lateral geniculate nucleus (LGN), ganglion cell, visual cortex, receptive field, fovea, lateral inhibition, diopters, focal length, troland, V lambda curve, Purkinje shift, mach bands.

Luminance: measurement of light intensity relative to human response to light Radiance: measurement of light intensity in terms of power produced, not always visible Hyperopia: farsightedness, eye might be too shallow (not enough power) Myopia: nearsightedness, eye might be too long; caused by close work or insufficient light Bits: binary digit in terms of a display that has two intensity levels (e.g. 2^10 bits = 1024 different states) Blind sight: lesions in primary visual cortex/V1 which causes "blindness"; people can still respond to visual stimuli Blind spot: spot in periphery lacking photo receptor cells caused by optic nerve; brain fills in information so we don't notice this Lateral geniculate nucleus (LGN): we don't really know what this does; receives sensory input from retina and relays information to primary visual cortex (connection for optic nerve to cortex) Ganglion cell: center-surround theory (on-center and off-center cells); neurons (Magno [respond faster and dominate periphery] and Parvo) that relay information from retina to brain; causes grid illusion Visual cortex: receives information from LGN; various levels, feed forward and feedback connections; right side of brain receives information from the left eye and vice versa Receptive field: particular region of sensory space in which stimulus will modify firing of neuron; measured in the world w/ electrode in brain and light in front of animal, listening to when it fires Fovea: depression in retina where visual acuity is highest with a high concentration of cones; center of field of vision focused here Lateral inhibition: inhibition that neighboring neurons have on each other - fire more strongly if neighbors don't (grid illusion); receptive fields in periphery are larger (illusion there and not in center); for on-center cells, more negative space stimulated Diopters: unit of magnifying power of lens (1/Focal Length OR 1/Object Distance + 1/Image Distance) Focal length: distance at which objects at infinity are focused Troland: measure of luminance; 1 candela/meter^2 OR 1 nit seen through a 1 mm^2 pupil (common for finding units of intensity) V lambda curve: visibility as a function of wavelength; scotopic light shifted left to bluer colors and photopic light shifted right to brighter colors Purkinje shift: peak luminance intensity shifts towards blue at night (lambda curve shifts); red flowers appear bright during day but black at night Mach bands: receptive field (center-surround mechanism) gives you some response and when you move right, there is less inhibition and the neurons fire more giving a slight increase before it falls again; removes inhibition at an edge and creates a little bump [false shadow on edge between light and dark object caused by lateral inhibition]

24. Explain why a negative lens helps a myope to see better.

- Eye is too long and image falls short of retina when object is far away (image is focusing too far forward) - Negative lens decreases extra power and increases focal length by diverging rays of light so they focus further back in the eye, at the retina

19. In terms of the power of the lens and cornea, describe the difference between someone near-sighted (myopia) and far-sighted (hyperopia). Describe how lenses help these people.

- Myopia: too much power (eye is too long and image falls short of retina when object is far away) - Presbyopia: lens hardens with age and there is no known cure - Hyperopia: not enough power (eye is too short and the image focuses behind the retina if the object is near) Concave lenses are used for myopia -Spread light out before it reaches the convex lens in the eye, therefore letting the image focus directly on the retina Convex lenses are used for hyperopia -Focuses light in addition to the natural lens and makes the image focus on the retina

28. What is astigmatism?

- Occurs when the cornea (clear front cover of the eye) is irregularly shaped or sometimes because of the curvature of the lens inside the eye and causes blurred vision - An irregularly shaped cornea or lens prevents light from focusing properly on the retina, the light-sensitive surface at the back of the eye. Vision can become blurred at any distance, leading to eye discomfort and headaches - Frequently occurs with myopia and hyperopia (refractive errors)

62. Name 4 reasons we may want to convert a non-visual data set into a visual data set. (W)

1. Ability to comprehend large amounts of data 2. Perception of emergent properties that were not anticipated 3. Enables problems with data itself to become immediately apparent 4. Facilitates understanding of both large scale and small scale features of data 5. Facilitates hypothesis formation

29. Why is skin so hard to model in computer graphics (two reasons).

1. Bidirectional Reflection Function (BRDF): Need to use this because skin has complex reflectance function. Different layers of skin reflect light differently and sense we are very familiar w/ skin, we are very sensitive to any faults. - BRDF is important because (1) many surfaces (e.g. faces) have complex reflectance functions and (2) the human visual system uses this function to determine material properties 2. If skin is too realistic in computer graphics, it approaches the Uncanny Valley

16. Plot the relationship between scotopic and photopic spectral sensitivity and use this plot to explain the Purkinje shift and why red light or red goggles are often used under scotopic conditions. (Look at graph.)

1. Draw wavelength x threshold graph 2. If you go up you saturate the rods (curve shifted up) [505 nm threshold for rods] 3. Cone sensitivity comes between saturation curve and rod curve [555 nm]; less sensitive than rods and shifted to the right There is a region where rods are not saturated, but cones are stimulated - use red goggles to maintain night vision and still be able to see. If in that region doesn't take make time to adjust to dim light.

83. Describe 5 ways to make a spot appear to move.

1. Move the spot 2. Adapt to motion then observe spot (i.e. adjustment to speed in car). 3. Auto-kinetic: observe dot without references 4. Induced motion: move reference frame (change the reference/background and the object may appear to move [e.g. moon or star in clouds] 5. Stroboscopic motion: flash successive dots (in different locations) Motion aftereffect: Changes the balance between neurons selective to motion in the directions away from and in the direction of the adapting motion. Can result in an under-estimation of speed. Can result in errors in direction of motion.

66. Describe the three primary stages of perceptual processing according to Ware. (W)

1. Parallel processing to extract low-level properties of visual scene - Rapid parallel processing - Extraction of features, orientation, color, texture, and movement patterns - Transitory nature of information, which is briefly held in an iconic state - Bottom-up, data-driven model of processing - Serving as the basis for understanding the visual salience of elements in displays 2. Pattern perception - Slower serial processing - Top-down attention being critical to the formation of objects and patterns pulled out from the feature maps - A small # (1-3) of patterns becoming "bound" and held for a second or two under top-down attentional processes - Different pathways for object recognition and visually guided hand motion (the perception and action channels) 3. Visual working memory Direct from textbook: - Stage 1: information is processed in parallel to extract basic features of the environment - Stage 2: active processes of pattern perception pull out structures and segment the visual scene into regions of different color, texture, and motion patterns - Stage 3: the information is reduced to only a few objects held in visual working memory by active mechanisms of attention to form the basis of visual thinking

49. For a 1.5-meter television screen at 1 meter, how many lines are required to be at the highest visible resolution (assume 100 PPD)?

1. START w/ A TRIANGLE 2. Might be asked about theta, distance, or size (fill in triangle) 3. Often want 100 pixels/degree tan(theta) = size/distance tan^-1(1.5/1) = 56 degrees Now we know theta, but how many lines do we want across the display? MULTIPLY BY 100 to get 5600.

45. How does one measure the contrast sensitivity function (CSF) of a human observer? Describe the contrast sensitivity function differences in adults and a 3-month old. What does this imply for information presented to children on a display?

1. Show a grating of different frequencies and look at how much contrast you need to see it 2. Reduce contrast - how different bright and dark regions are until you just can't see it -Spatial frequency x sensitivity -Sensitivity is also 1/threshold -Given low frequency and reduce contrast until we can't see it 3. Measure threshold and discover you need very little contrast at some frequencies and plot as sensitivity and take inverse of it (spatial frequency); peaks at about 4 cycles/degree [look up what actual graph of contrast sensitivity function looks like] Much smaller for 3-month old Children are picking things up at much lower frequencies - things farther away lose detail. So when doing studies, must consider those sensitivities.

63. Describe the four stages of information visualization. What is the role of feedback in these stages? (W)

1. The collection and storage of data. 2. A preprocessing stage designed to transform the data into something that is easier to manipulate. Usually there is some form of data reduction to reveal selected aspects. Data exploration is the process of changing the subset that is currently being viewed. 3. Mapping from the selected data to a visual representation, which is accomplished through computer algorithms that produce an image on the screen. User input can transform the mappings, highlight subsets, or transform the view. Generally this is done on the user's own computer. 4. The human perceptual and cognitive system (the perceiver). The longest feedback loop involves gathering data. Another loop controls the computational preprocessing that takes place prior to visualization. Feedback in these stages allows for a different vantage point or to better understand emerging structures.

47. Use a space time plot to explain why directional blur is sometimes introduced into movies or games with fast motion. Why don't we notice the blur?

1. Time on y-axis and space on x-axis 2. Move at different times to different positions Setting same frame up three times - jumping movement when something is moving; can blur in direction of motion creates creates smoother motion and blurry images. This would normally be a problem if looking right at it, but is fine in the background. Sampling artifacts evident unless we blur in direction of motion. Even high sampling rate can't fix it. Worse with faster motion. Acuity drops with retinal velocity. Blur also doesn't happen where person is supposed to be focusing. Three solutions to sampling problem 1. Speed up projection rate: problem - helps flicker but not fast motion 2. Speed up filming rate and project rate - problem: requires more film and new technology 3. Spatio-temporal anti-aliasing -With frame rates low relative to velocity, sampling artifacts become visible -Acuity drops with retinal velocity -Blurring in direction of motion (e.g. similar to leaving the shutter open) can reduce the sampling artifacts. This involves considerable more work in computer graphics.

48. For a 1-meter wide television with 4000 pixels across, how far away should it be, to be at the highest visible resolution (assume 100 pixels/deg PPD)?

100 PPD is the gold standard for resolution 4000/theta = 100, theta = 40 degrees How far back do we need to be to get a visual angle of 40 degrees? 1.37m

34. Provide a simple map of the pathways that take visual information from the eye to the brain. Include a discussion of the left and right visual fields and the superior colliculus.

90% of optic of LGN projects to the cortex and 10% to structures like the superior colliculus - Opposite sides mapped - Superior colliculus: direct behavioral response to visual stimuli (i.e. if someone hits a baseball you naturally follow it with your eyes) Start w/ two eyes and two halves of the brain: right visual field projects onto left side of BOTH eyes to the left side of brain; left visual field projects onto right side of BOTH eyes to right side of brain -> goes through LGN and some information goes off to the superior colliculus

50. For virtual reality glasses with a visual field of 180 degrees, how many pixels across are needed to be at the acuity limit? If the display is binocular with 120 degrees in each eye, how many pixels are required for each eye?

Acuity limit = 50 cycles/degree (100 lines/degree under ideal circumstances) 180 * 100 PPD = 18000 pixels 120 * 100 PPD = 12000 pixels/eye

77. How does the visual system factor out the amount and color of the illumination when making lightness decisions? (W) (Won't ask about color in this exam.)

Adaptation: changing sensitivity receptors and neurons in eye based on environment's illumination Bleaching of photo receptors: less sensitive in bright light, regain sensitivity in less light - opening and closing of iris Simultaneous contrast: looking at lightness relative to background Lateral inhibition: as brightness increases, neurons are stimulated more but also inhibited more creating perception of constancy Additional: 1. Brain must somehow take the direction of illumination and surface orientation into account; a flat white surface turned away from the light will reflect less light than one turned toward the light 2. Brain seems to use the lightest object in the scene as a kind of reference white to determine the gray values of all other objects 3. The ratio of specular and nonspecular reflection to differentiate black from white

37. Why is accommodation a problem for virtual reality? Why is this not a significant problem for those older than 60?

All current virtual reality headsets have a fixed focal length. This can create problems for those younger than 50 because they try to accommodate to the appropriate distance. Those older than 50 can't accommodate.

68. What are augmented reality displays? Describe three practical uses. (Won't be on THIS exam.)

Augmented reality: superimposing visual imagery onto real world so that people see computer-graphics enhanced view of the world Applications: 1. Radiologist sees inside woman's breast tissue to guide biopsy needle 2. Automobile servicing: technician can see instructions & structural diagrams superimposed on actual car 3. Military application: pilot can see enemy/friendly locations superimposed on view of landscape

84. Be familiar with the following concepts: Autokinetic motion, induced motion, apparent motion, motion aftereffects, persistence, low and high thresholds for motion. From the online readings: Divided Attention by David Glenn.

Autokinetic motion: In dark without references, small lights may appear to move. Induced motion: Object appears to be moving because of other objects moving nearby in the visual field. Could be moving background. Apparent motion: An optical illusion in which stationary objects viewed in quick succession or in relation to moving objects appear to be in motion. Short range apparent motion: At short distances and fast frame rates, apparent motion is equivalent to real motion. The limits of temporal and spatial resolution makes them indistinguishable. Long range apparent motion: Correct time between flashes far apart (60 ms) creates illusory motion. Motion aftereffects: Visual illusion after viewing moving visual stimulus and then fixating on stationary stimulus. Stationary stimulus appears to move in opposite direction. Caused by motion adaptation. Persistence: Visual stimulus persists even after it is turned off. Depends on total amount of light (time*intensity). Explains why we can't sense temporal flicker with lights. Thresholds for motion: With reference can detect much smaller amounts of motion. - Low: (How slow can something move where we can still detect movement) No reference = 10-20 minutes of arc/second; Reference = 1-2 minutes arc/second - High: (How fast can something be moving where we can still detect it) Full acuity = up to 2 degrees/second. Acuity drops with increasing frequency

36. Describe the roles that the cornea and lens play in focusing an image.

Background: - Cornea: fixed power of ~40 diopters - Lens: variable power of 10-22 diopters (at young age) The lens focuses light through the vitreous humor, a clear gel-like substance that fills the back of the eye and supports the retina. The retina receives the image that the cornea focuses through the eye's internal lens and transforms this image into electrical impulses that are carried by the optic nerve to the brain. - The cornea functions like a window that controls and focuses the entry of light into the eye - By changing its shape, the lens changes the focal distance of the eye - it focuses the light rays that pass through it (and onto the retina) in order to create clear images of objects that are positioned at various distances

40. Describe 2 similarities and 2 differences between the processing of objects along the visual pathway and the processing of objects by deep networks.

Background: - Deep learning (also known as deep structured learning, hierarchical learning, or deep machine learning) is a branch of machine learning based on a set of algorithms that attempt to model high-level abstractions in data by using a deep graph with multiple processing layers, composed of multiple linear and non-linear transformation. - Deep neural network: 1. Begin w/ a massive amount of labeled data 2. Feed that into the network 3. Check if it gave you the right answer 4. Correct the network by "back propagating" the errors - This is NOT the way brains learn. Visual systems are much the same across primates: learn w/out labeled data. Similarities: - Both systems are fairly hierarchical (multiple-range) - Very similar in early stages (e.g. first stages detect edges) Differences: - Deep networks 1. Supervised 2. Massive amounts of labeled data 3. Requires thousands of examples 4. Feed forward after learning - Mammalian visual systems 1. Largely unsupervised 2. Very little labeled data for learning 3. Requires a handful of examples 4. Massive feedback recurrent

31. Provide two examples that demonstrate that shadows play a role in the interpretation of objects.

Background: - For a given surface, an observer must decide how much of the luminance is due to the illumination and how much is due to the reflectance Examples: 1. W/out shadow teapot looks like it's floating in space (shadows can help determine the location of an object: attached v. floating) 2. Changing shadow makes basketball look like it is bouncing, sliding, floating, etc. 3. Google logo: shadows can help determine the layout of an object

55. Explain the contrast sensitivity function under scotopic, mesopic and photopic conditions. (Make sure you know the graph.)

Background: - How far away can I recognize someone's face in daytime? Not good reliable eyewitness if saying they recognized someone across from a parking lot at 7pm - The contrast sensitivity function has the potential of adding more information about the functioning of the visual system than that given by visual acuity, because it assesses sensitivity over a wide range of spatial frequencies, while visual acuity measures primarily sensitivity at the high spatial frequencies Acuity and sensitivity drop as you go into darkness (photopic -> mesopic -> scotopic)

11. Describe four differences between foveal and peripheral processing (how the information is processed). (E.g. If building a display, how would you take advantage of foveal v. peripheral processing.)

Background: - Peripheral: rods; scotopic vision - Foveal: cones; photopic vision 4 differences: 1. Light sensitivity - peripheral better in dim light, foveal better in bright light 2. Wavelengths - peripheral sees blue-green better (peak at 505 nm; Purkinje Shift), foveal sees color better (peak at 555 nm) 3. Acuity - peripheral low in acuity, foveal high in acuity 4. Flicker response - peripheral more responsive to flicker although rods are actually slower, foveal less responsive to flicker E.g. In an animation, have background images be lower resolution since acuity is lower in the periphery

8. Under dim conditions (e.g. finding a star at night), it can be useful to look two or three degrees off the target. Why? (See figure in textbook.)

Background: - Rods: scotopic vision; peripheral - Cones: photopic vision; foveal Looking right at the star uses the fovea, which would make it too dim. Looking two or three degrees off the target activates the rods enough so that acuity doesn't drop; place the image of the object on the part of the retina that is rich in rods.

2. What will be perceived as brighter? A 100-watt lamp that produces most of its energy in the range of 500 to 540 nanometers or a 100-watt lamp that produces most of its energy between 650 and 700 nanometers?

Background: - Scotopic vision: dim light (peaks at 500 nm) - Photopic vision: bright light(peaks at 550 nm) Answer: - 500-540 nm light because photopic vision peaks at 550 nm - Radiometric (watts): physical, total energy, whether visible or not If question was asked w/ nits: - Both would be perceived as having the same brightness - Photometric (candelas/meter^2 = nit): takes visual sensitivity into account

9. Describe two aspects of virtual reality that can make us sick.

Background: - Simulator sickness: Conflicting cues from visual system and vestibular system of inner ear; causes nausea. Don't expect participants to move heads from side to side while wearing a helmet-mounted display. Contributing Factors: 1. Active conflict - action does not equal vision; induced self-motion not in agreement between visual and vestibular systems 2. Update rates not high enough (slow computers) - slower than what the brain processes and causes a discord between processing and refreshing rate 3. Poor resolution - doesn't keep pace with user's head movements (expectations violated) 4. Frame rate 5. Gender (women more likely) 6. Experience w/ VR

33. Provide a rough estimate of the spatial frequency of the grating shown on the blackboard (in cycles/deg) and explain how you arrived at this estimate. Approximately, how many lines would there need to be to approach your acuity limit? (Also tell us where you are sitting.)

Background: - Spatial frequency: acuity decreases for high-frequency patterns as you increase distance. Also depends on high contrast vs low contrast. - Acuity limit: around 50 cycles/degree, or 100 cycles/degree under best conditions For a 4K display, at what distance do you need to sit for it to be at your acuity limit: - 4000 pixels/X degrees = 100 pixels/degree - x = 40 degrees - tan(theta/2) = (size/2)/distance tan(20) = (size/2)/distance distance = size/(2*tan(20)) - For a 1 meter screen distance = 1/(2*0.36) = 1.38 meters Actual answer (maybe): - 1 thumb width = 2 cycles/degree - Acuity limit = 50 cycles/degree, which is 100 lines (2 cycles/degree) = 200 lines to reach acuity limit

18. Suppose that you have a near point of 10 cm and a far point of 25 cm. A. What is your condition? (e.g. hyperope?) B. What is your amplitude of accommodation (the diopter range of the lens/flexibility of diopters)? C. How should this person be corrected? D. After correction, where will this person's near point and far point be located?

Background: - np = 0.1 meters -> 1/0.1 = 10 diopters of EP - fp = 0.25 meters -> 1/0.25 = 4 diopters of EP - At about 20 you have 12 D and at about 60 1 D; person is 40 A. Myopic, but also presbyopic because range is going down B. 10 D - 4 D = 6 D C. We want far point to be at infinity. -4 D lens to correct. (Aim for 0 D because 1/0 = infinity.) D. After correction, have accommodation range of 0 to 6, gives far point at infinity, near point at 1/6 D = 16.7 cm. This person can't see beyond 0.25 D (concave lens w/ 4 D too much power; give -4 D lens to get to 0). - NEW NP = 1/6 D = 0.17 cm - NEW FP = 1/0 D = infinity

80. Ware uses the concept of 'Brain Pixels' to describe the efficiency of a display. What is a "Brain Pixel" and how might it be measured? What is an efficient system by his measure? What typically happens to efficiency with increasing display size?

Brain Pixels: image units used by brain to process space - Retinal ganglion cells best capture the brain pixel idea - Receptive field size = 0.006(e + 1) At 10 degrees of eccentricity, w/ the big screen there is an approximate match between screen pixels and brain pixels. Even though a conventional monitor covers only about 5 to 10% of our visual field when viewed normally, it stimulates almost 50% of our brain pixels. This means that even if we could have very high-resolution large screens, we would not be getting very much more information into the brain.

60. Why do we see colors on the surface of a CD? (L)

CD's have little grooves on the back of them that cause diffraction. Different colors of light have different wavelengths and therefore diffract differently. Colors change depending on angle relative to viewer.

30. How does the early visual system compute ratios? Why is this a good thing?

Center-surround of Ganglion cells: inhibited and excited, compute ratios with log (Ic) - log (Is) = log (Ic/Is) This is a good thing because: - Cones have compressive relationship (log of intensity). By having cones behave w/ compressive response create ratio. - Ratios help preserve light constancy. Ratios will stay the same even with added illumination.

69. What is chromatic aberration and what is its cause? What is an illusion that arises from this property of the eye? (L) [REVIEW TEXTBOOK AND ELABORATE ON ANSWER.]

Chromatic aberration: different wavelengths of light are focused at different distances in the human eye Blue text on red background/ Red text on blue background Mixture of people see red closer, blue closer, or on same plane

52. What is dithering? Why is it effective perceptually?

Dithering: a strategy when the number of intensity levels is not high, but the resolution is high Takes advantage of human eye's tendency to mix two adjacent colors. Number of intensity levels is not high but resolution is.

59. Why do we see colors in an oily puddle? (L)

Due to difference in thickness of oil across surface of water, light rays travel different distances (some reflect immediately off of oil, some refract), mix together, and produce different colors. Constructive & destructive interference

6. Explain why the dynamic range of a display depends on the amount of stray light. Provide an example.

Dynamic range: ratio of brightest pixels to darkest pixels. Light reflected back onto the screen decreases the dynamic range. Thus, not always good to have very bright and very HDR displays. E.g. Screen w/ 1,000:1 range. 1% light reflected back onto screen. Increases range to 1,010:11 = 100:1.

73. Provide a brief explanation of the Cornsweet illusion. (W and L)

Edge enhancement: Lateral inhibition can be considered the first stage of an edge detection process that signals the positions and contrasts of edges in the environment. One of the consequences is that pseudoedges can be created; two areas that physically have the same lightness can be made to look differently by having an edge between them that shades off gradually to the two sides. The brain does perceptual interpolation so that the entire central region appears lighter than surrounding regions. This is called the Cornsweet effect. In the Cornsweet illusion, the region adjacent to the light part of the edge appears lighter, and the region adjacent to the dark part of the edge appears darker. In fact, they have the same brightness.

1. Describe the differences between efficient, inefficient, and restrictive (limited) display systems. Do we need all displays to be efficient?

Efficient - information matches limits of sensory system (e.g. Apple retina displays, anti-aliasing) Inefficient - more information than the sensory system can handle (e.g. Japanese train schedules) Restrictive (limited) - below the sensory limits of the visual system (e.g. stop sign) No, because we have limitations in capacity: - Reception (the window of visibility); wavelength, intensity, spatial resolution (pixels/degree), flicker (temporal resolution) - Discrimination; intensities, colors, spatial resolution, flicker - Capacity; how much information can be handled quickly and efficiently (7 plus or minus 2)? - Biases and interpretation; facial recognition

3. Calculate the visual angle of your thumb at arm's length, your big toe standing up, and the moon. Show your calculations.

Equation (large field displays): tan(theta) = size/distance theta = tan-1(size/distance) Equation (small field displays): tan (theta/2) = (size/2)/distance Step 1: measure arm length (D), and thumb length (S); triangulation Step 2: count # of thumbs across Step 3: calculate, where theta is give in degrees (2 to 3.25) Example: Square w/ lines is about 2 degrees across (measured with thumb). How many lines required to be at resolution? 2 x 100 = 200

46. Why are sine waves often used when describing a linear system?

Fourier Transform/Fourier Series - https://en.wikipedia.org/wiki/Fourier_series

38. What is foveated rendering? What are the advantages over standard rendering?

Foveated rendering: technique to put high resolution rendering (graphical detail) where the observer is looking; in foveated rendering, only have high definition in thumb area Reduces computations needed for rendering in VR but requires that the eyes can be tracked

4. What is the gamma of a display? Roughly what is the gamma of your visual system? How does the gamma of your visual system affect the apparent luminance of an intensity ramp?

Gamma: distribution of its intensities, given by output = input^gamma (luminance = voltage) The visual system has a gamma of about 0.5 Gamma turns intensity graph into a nonlinear square root function. This compressive non-linearity reduces range of responses needed to represent an input w/ a large range of intensities (brightness = intensity^0.5). In other words, in bright regions, the visual system needs larger intensity changes to be visible. These large step sizes allow for good compression (digital images). Images are then displayed w/ a gamma of 2 to allow for a linear relation between input and output. Monitor gammas are intentionally non-linear to cancel out visual system gammas to make the monitor luminance appear to change linearly.

7. What is high dynamic range (HDR) imaging? How does this approach increase the apparent range of an image?

HDR imaging - capturing HDR images and creating images that look HDR on LDR displays (take pictures w/ various exposures and then use tone mapping to re-map luminance to screen's range, maintaining contrast across various areas of the image) HDR display has a huge range - efforts to stretch out apparent range by increasing contrasts can sometimes make images appear garish; can also increase range of luminances produced by display 1. Creating images w/ higher dynamic range 2. Displaying those images on displays w/ high dynamic range 3. Displaying those images on low dynamic range displays - but making them appear more like high dynamic range

87. What is a switching cost - and how does this apply to computers in the classroom?

Heavy multi-taskers spend important resources switching between sources and monitoring those sources. For example, the mere presence of a cell phone or computer has been shown to diminish attention and reduce performance on cognitively complex tasks (single-minded attention is vital to learning). E.g. letter-recitation task to number-recitation task

74. Explain the illusions created by the Hermann Grid and Chevreul/Mach bands in terms of receptive field spacing. Diagrams would be useful here. (W - 73/74)

Hermann Grid: more inhibition at the spaces between two squares, so they seem brighter than the regions at the intersections; the black spots that are seen at the intersections of the lines are thought to result from the fact that there is less inhibition when a receptive field is at the corner of four squares than between the edges of two squares (DoG model of concentric opponent receptive fields) Mach Bands: at the point where a uniform area meets a luminance ramp, a bright band is seen; appear where there is an abrupt change in the first derivative of a brightness profile Chevreul Illusion: when a sequence of gray bands is generated, the bands appear darker at one edge than the other, even though they are uniform; can be simulated by the application of a DoG model of the neural receptive field

35. Distinguish between hypermetropia, myopia, and astigmatism including the symptoms, causes, and remedies of each. (Answers similar to some above.)

Hypermetropia/Hypermyopia - Symptoms: not being able to see things up close - Causes: eye is shorter than normal, cornea not curved enough, lens sits too far back; doesn't have an environmental cause, so genetic - Remedies: glasses with positive correction Myopia - Symptoms: not being able to see things far away - Causes: eye elongated; close up work, not enough bright light, genetic predisposition - Remedies: get progressive glasses to slow down progression, get negative corrective lenses Astigmatism - Symptoms: distorted vision due to eye being misshaped, light rays scattered - Causes: irregularly shaped cornea/lens - Remedies: corrective lenses/refractive surgery

21. What is the relation between lens flare and perceived brightness. Why does this work?

Imperfections in lens creates a spray of light that the visual system uses as a cue to determine brightness by measuring how wide the flare is; more lens flare = brighter world This works because we are not good at perceiving high levels of brightness intensity, so instead we perceive beyond a certain threshold as the light spreading out/being wider, hence the lens flare

65. Provide an example of a stage in visual processing where there is considerable compression (hint the optic nerve). What is the magnitude of the compression?

In actual exam may not give hint The compression is in the optic nerve. About 120 million photoreceptors in each eye. About 1 million fibers in optic nerve. 120 million photoreceptors compressed to 1 million; losing resolution in periphery. 120:1 neurofibers. Optic nerve to cortex start to see sparse representation instead ( a lot of neurons but few are active). The magnitude of the compression is 120:1

67. What is the simplest model of surface texture that creates the perception of surface shape from shading? How is this used to shade a sphere with one light source? What are two additional effects that improve how the shading looks? (W)

Lambertian shading: if we have a perfectly matte surface, how bright the surface appears depends only on the cosine of the angle between the incident light and the surface normal Lambertian shading can be used w/ specular, ambient, and cast shadows - Specular: the light that is reflected directly from a surface (highlights on glossy objects); angle of reflection = angle of incidence - Ambient: ambient light is the light that illuminates a surface from everywhere in the environment, except for the actual light sources - Cast: an object can cast shadows either on itself or on others objects; can greatly influence the perceived height of an object

23. What role does lateral inhibition play in lightness constancy?

Lightness constancy: observers can usually identify reflectance (brightness) of material despite changes in illumination which helps visual system factor out color and amount of illumination and signals differences in light levels especially at edges of objects - As overall illumination increases, ganglion cells more excited but also more inhibited by neighbors, creating constant perception of brightness - When looking at a uniform field, receptive field measures difference between center and surround (ratio) rather than absolute values and this ratio stays the same

25. What is presbyopia? Can you avoid it? What does this do to the near and far points?

Loss of lens accommodation with age Can't avoid it Near point increases, far point decreases

44. What is the difference between lossless and lossy compression? What role does the visual system play in these compression schemes?

Lossless: removes redundancy but loses no information (allows for perfect reconstruction) -Any form of prediction in a data set allows for some lossless compression Lossy: removes structure that is hopefully invisible to the visual system Optimal compression must match the statistical structure of world and match image coding ability of observer. Better compression achieved if you can accept some (hopefully non-visual) loss. Two reasons a compression algorithm can be successful: 1. The algorithm takes advantage of the redundancy of the data 2. It sends only the information used by the observer

81. Do a news search for "eye tracking" (e.g., google news). Describe 3 recent applications of eye tracking.

Methods of tracking the eyes: 1. Contact lens 2. Infrared sensing (head mounted) 3. Electro-oculograph -Records eye movements by measuring small electrical charges with tiny electrodes attached to the skin at the inner and outer corners of the eye (sleep research, low accuracy, but unobtrusive) 4. Image processing techniques -Allows head movements. Typically overlays image with eye movement. Recent applications: 1. Medicine - reveal abnormalities in eye functioning and to conduct cognitive studies focused on learning about peoples' emotions and intentions 2. Retailer H&M is using the new Tobii Pro Sprint to better understand how online shoppers navigate their website. Set up participant w/ program and use it to record participant's eye movements. The recording shows where the user's gaze falls, potentially revealing which design features attract their attention, which they ignore, and which ones confuse them. 3. Piloting a drone - system can understand both the location and orientation of the drone and its pilot (takes instructions based on user's orientation). Tobii Pro's Glasses 2 used to track eye movements of pilot. Glasses are plugged into an NVIDIA Jetson TX2 CPU and GPU and deep neural network takes the incoming images from the glasses, crunches the numbers, and is able to calculate how far away the drone is based on its perceived size. From there, it's just a case of gazing at your chosen location and the glasses will translate that data into a vector for the drone.

39. Be prepared to identify the following: micro saccades, tremor, saccades versus smooth movements, vestibuloocular reflex, optokinetic nystagmus, dynamic acuity.

Micro-saccades: Small, jerk-like involuntary eye movements. Occur during prolonged fixation. Could correct eye drifting (fixational eye movement). Tremor: Constant movement in all healthy people. Due to constant activity in brain stem (fixational eye movement). Saccades v. Smooth Movement: Smooth has lower latency (150 ms vs 250 ms). Need something to track for smooth movement. Might need saccades to catch up. Smooth pursuit is not possible without tracking. Vestibuloocular Reflex: Activation of vestibular system causes eye movement. Stabilizes vision. When you turn your head, eyes move in opposite direction. -Vestibular system feeds into regions of the brain controlling eye-movements -Movement of the head produces change in eye position to allow stabilized vision -Some antibiotics affect the vestibular system - very difficult to walk -E.g. nausea from glasses w/ new prescription -E.g. motion sickness + Oculus Rift: if display has a slow response, there can be a violation of expectations of where your eyes will be when your head moves Optokinetic Nystagmus: Involuntary response to movement. Drifting grating produces saw tooth response and forced fixation produces smaller saw tooth response. Dynamic acuity: Visual acuity while smoothly pursuing object (e.g. following baseball).

42. Why doesn't an image normally disappear if I fixate very accurately at one point? What does one have to do to make it disappear?

Neurons in visual system only change in order to keep firing. If they get exactly the same stimulation in the same pattern, they will stop firing. You cannot make your eyes stay still - will normally give small movements or tremors (edge). You need to stabilize something very accurately on the retina (e.g. retinal blood vessels - don't see them because they move with the eye) or use a blurry object with small movements that don't create a large change in the image and thus doesn't produce much change to neurons.

64. Are abstract (arbitrary) visual representations based only on cultural knowledge? Explain. (W)

No: - Deregowski: people in Zambia able to match pictures of toy animals to real animals - Hochberg and Brooks: raised daughter without pictures/social input as to what pictures represented. She was still able to identify line drawings. - Arbitrary symbols - if you are outside of the culture, can't recognize what "dog" means, but if using pictures, is cross-cultural. Entirely learned. No relations between the symbol and what it represents. The word "dog" and a dog have no relationship. - Sensory symbols - takes advantage of visual processes and is cross-cultural.

51. Describe the steps for producing anti-aliasing on a computer screen. Why is this done and what does it achieve?

Option 1: 1. Compute average of light pattern that each light pixel represents. 2. Take high resolution, blur it, subsample. 3. Slowly shift mean intensity Option 2: 1. Take high resolution images 2. Blur (remove frequencies before they are aliased) 3. Subsample Achieves higher-than-device resolution; made up of a combination of foreground and background colors. Solution when the display has many intensity levels but not sufficient resolution.

26. Describe three cues that observers use to discriminate lightness from brightness.

Penumbra: Light sources are not point sources. Shadow produces a little bit of blur at the edge and that is the penumbra while the dark shadow is called the umbra. Multiplication across borders: Ratios of intensity maintained along illumination borders. For a given surface, an observer must decide how much of the luminance is due to the illumination and how much is due to the reflectance. If the ratio of luminances is constant across the luminance boundary, we perceive this as an illumination boundary. Co-planar hypothesis: Surfaces in the same plane are assumed to have the same illumination without other cues that signal a shadow. The "perceived" illumination (brightness) changes the interpretation of the perceived lightness (reflection).

76. Describe the relationship between stimulus intensity and intensity of perception that Fechner and Stevens found. How does this relate to the monitor gamma? (W)

Popularized technique known as magnitude estimation to provide a way of measuring the perceptual impact of simple sensations - S = al^n - This law states that perceived sensation S is proportional to the stimulus intensity I raised to a power n. - Perceived brightness = luminance^n The relationship of physical luminance to the input signal on a monitor is approximated by a gamma function: L = V^gamma where V is the voltage driving one of the electron guns in the monitor, L is the luminance, and gamma is an empirical constant that varies widely from monitor to monitor. This monitor non linearity is not accidental; it was created by early television engineers to make the most out of the available signal bandwidth. They made television screens nonlinear precisely because the human visual system is nonlinear in the opposite direction, as Stevens had observed.

5. Roughly, for a single image, what is the range of intensities found in the real world, modern LCD televisions, and hard copy?

Real world - 300:1 indoors; 30,000:1 outdoors Modern LCD televisions - 5,000:1 Hard copy (oil paintings and drawings) - 30:1

43. Describe the five different types of eye-movement and an example of when each is used (or describe the eye-movements that occur as you are running around your house trying to catch your cat - or playing tennis). (See question 39.)

Saccadic: scanning a page Smooth Pursuit: following a baseball Fixational (Tremor, Micro-saccades, Drift): staring at something Vestibular-Ocular Reflex: eye movement when you turn your head side to side Optokinetic Nystagmus: following scenery outside of car window

75. What is the Cornsweet illusion (a picture could be useful)? How has this effect been used by painters? (W)

See 73 Use by painters to enhance edges, making objects more clearly distinct given limited dynamic range of paint

13. What are the conditions that produce the appearance of self-motion for a human observer? What was wrong with the drum studies? How would you design a perceptually 'efficient' automobile simulator if the observer limited his/her direction of view to straight ahead? (What do you want to happen in the fovea that doesn't happen in the periphery?)

Self-motion: 1. best - linear motion in periphery 2. works - radial motion in fovea 3. does not work - radial motion in periphery or linear motion in fovea Drum study - central part of nuclear field doesn't make you feel like you're moving, but periphery does. Didn't stimulate normal movement - missing radial motion, only gave linear motion. Have simulator project linear motion in periphery. Can also have low resolution and sloppy color in periphery.

58. What does it mean when we say that the visual system uses a "sparse" representation?

Sparse representation: small # of active neurons w/ strong activation/action For many kinds of algorithms, few neurons responding - this does happen in the optic nerve, but goes back up again in the cortex. They are not minimizing how many neurons there are, but minimizing how many neurons are active to action. How many pixels to represent an edge? Cortex has edge detector and doesn't need other neurons to respond. What that does is, is it produces a large # of neurons and a small # of ACTIVE neurons.

32. What is a specular reflection? What is a diffuse reflection? How can these two features combine to affect the perception of a material?

Specular reflection: screen pointed at something shiny like a phone, has high specular reflection (same direction - glossy) Diffuse reflection: when light comes at surface from one direction, will go out in all directions (evenly - matte) Phong Reflectance: pure specular will not give you a reflectance except where light in = light out; can create a combination of two features to create sense of something like a billiard ball (soft to relatively reflective material) - Early Toy Story films show a very plastic world using simple reflectance models

41. Describe the difference between supervised and unsupervised learning.

Supervised: discover patterns in the data that relate data attributes w/ a target (class) attribute -These patterns are then utilized to predict the values of the target attribute in future data instances Unsupervised: data have no target attribute -We want to explore the data to find some intrinsic strategies in them

10. The famous blue and black (or white and gold) dress is seen differently by different people. Why?

The dress can be interpreted in two ways: black and blue under a yellow-tinted illumination OR white and gold under a blue-tinted illumination Judged as different forms of illumination (brain makes assumptions); perceived illumination (brightness) impacts perceived reflectance (lightness)

17. What is the Hermann grid and what is the current explanation?

The grid consists of black squares separated by white lines. You see faint black dots over each intersection of white lines. Illusory spots are perceived because the center/surround cells that signal the white of the intersections are more suppressed (by the four bits of white lines in their surrounds) than are the center/surround cells that signal the white of the lines between the intersections (because they have only two bits of white line in their surround). Thus, the cells responsible for signaling whiteness at the intersections are more suppressed, and therefore less active, than the cells signaling whiteness along the lines between the intersections, so the intersections look less white, and therefore darker, than the crosspieces.

85. What is the illusion of competence?

The illusion of competence - heavy multi-taskers are often extremely confident in their abilities, but they are actually worse at multitasking than most people.

22. Why would late night stores blast 17,000 hertz tones?

The mosquito tone upsets people who are younger; adults CAN'T hear this tone so it keeps kids from staying at these stores later because they CAN hear it

86. Are heavy multi-taskers good or bad at multitasking? Explain.

The multitasking illusion: chronic multi-taskers were terrible at ignoring irrelevant information, organizing information in their heads, and switching from one task to another. The illusion is that we feel that we are processing both streams of information, but we are actually processing less total information. The heaviest multi-taskers are the most easily distracted and perform worst in an environment with distractions. They spend important resources switching between sources and monitoring those sources. Illusion of competence

14. What causes motion sickness? How does this relate to the mismatch hypothesis? Why might motion cause sickness in the first place? Why is one more likely to get sick on a camel as opposed to a horse? Why is a passenger more likely to get sick than a driver even when they are both keeping their eyes on the road?

The standard theory is that when you are moving around with closed eyes, you have a sense of where you are (middle ear has semicircular canals w/ three angles which have a fluid that moves and stimulates neurons). Endolymph produces the sense of motion. A partial answer is the mismatch hypothesis - conflict between visual and vestibular systems. Importantly, sensory expectations are violated as well. Visual/vestibular errors are signs of poison Camels cause more sickness because they have ~2 Hz oscillation; horses have higher oscillation Driver causes motion and has expectation of when the motion is going to be, whereas passenger not only experiences mismatch between the visual and vestibular systems, but also doesn't have expectations of how they are supposed to move

20. Describe three recommendations for children to reduce their chance of getting high myopia? What has happened to the eyes of those that develop high myopia? Is myopia inherited or is it entirely environmental?

Three recommendations: 1. Don't do work too closely 2. Make sure there is sufficient bright light 3. Take preventative measures if you know you have a genetic predisposition towards myopia High myopia: - Eye is too long and image falls short of retina when object is far away - If objects are typically close, need more power in lens. Eye adjusts by continuing to grow, resulting in myopia (good for work up close) Inherited or environmental? Inherited: - Probability of child having myopia if they have myopic parents is 3x probability when neither parent is myopic - Scientists have pinpointed variation in APLP2 gene that correlates to five fold increased risk of myopia in childhood Environmental: - Nearsighted kids spend on average 3.7 fewer hours outside per week (less natural light) - Ultra-Orthodox Jewish boys who study Torah up close have higher rates of myopia than the girls who don't

12. How many "bits" of luminance are available with a display that has only two intensity values (i.e., a monochrome display). How many are needed to produce photographic quality images? What are three factors that have bearing on this number.

Two intensity values have 1 bit of luminance (2^1) Max is 1000 intensity levels (10 bits) to produce photographic quality images (smoother boundary levels) Three factors: 1. Screen range 2. Particular image - smooth intensity variations require more bits 3. Screen gamma

82. What is the uncanny valley? Provide a graph (label the axes). Describe the perception at different points along the curve. (Make sure you know graph.)

Uncanny Valley: when robots/other facsimiles of humans look and act almost like actual humans it causes a feeling of revulsion among human observers (effects made more dramatic if robot is moving)

71. What is cortical magnification? How does this relate to the eye chart of Anstis with letters of varying sizes as a function of eccentricity? (W)

Variation in acuity with eccentricity (measurement of distance from fovea in terms of visual angle) comes from cortical magnification. In the middle of the visual field, at the fovea, we can resolve about 100 points on the head of a pin. At the edge of the visual field, we can only discriminate objects the size of a fist. Cortical magnification refers to the fact that the number of neurons in the visual cortex responsible for processing the visual stimulus of a given size varies as a function of the location of the stimulus in the visual field. If you look at the center of the chart, each of the characters is equally distinct. Anstis took measurements of the smallest letter that could be seen at many angles of eccentricity from the fovea. In this version, each letter is about 5 times the smallest resolvable size for people with 20/20 vision.

79. Neurons are the basic unit of computation in the brain. How do neurons communicate with each other? Name one way in which they are similar to transistors, the basic unit of computation in a computer. (W)

Via electrical events called action potentials and chemical neurotransmitters. At the junction between two neurons (synapse), an action potential causes a neuron to release a chemical neurotransmitter. Neurons respond with discrete pulses of electricity. Neurons are constantly active, emitting pulses of electricity through connections with other cells. The rate of firing can be increased or decreased as the neuron is excited or inhibited.

27. What type of lens does a virtual reality display require if the display monitors are mounted 10 cm in front of the lens and are intended to be viewed at infinity? What type of lens is required if the observer normally requires -2 diopters of correction?

Want 10 D to put on it and make it look as if it is at infinity. What was at infinity before now looks like 0.1 m. No correction: 1/.1 + 1/infinity = 10 D lens. If you are already myopic (-2 D), can reduce by 2 D to 8 cm (-2 + 10 = 8). -2 D correction: 10 D - 2 D = 8 D. Should have an 8 D lens or just put glasses over VR display.

56. How can flicker above the critical fusion frequency still have an effect on a human observer?

Waving hand in front of face -> strobe (i.e. motion -> strobing) Critical fusion frequency: frequency at which flicker appears to be constant light source to observer. Related to visual persistence. Fluorescent lights above threshold, but can cause "building sickness". Incidence of headaches found to decrease when replacing 50 Hz fluorescent with 100 Hz.

78. Explain Weber's law regarding lightness differences. How much more luminance does a patch need to have to stand out from the background? (W)

Weber's Law states that if we have a background with luminance L and superimpose on it a patch that is a little bit brighter (L + δL), then the value of δ that makes this small increment just visible is independent of the overall luminance. Thus, δL/L is constant. Typically, under optimal viewing conditions, we can detect the brighter patch if δ is greater than about 0.005. In other words, we can just detect about a 0.5% change in brightness.

61. Describe the primary differences between the 'what' and 'where' systems along the visual pathway. (L)

Where (Dorsal): object's spatial location relative to viewer. Spatial awareness for actions (i.e. reaching). Contains detailed map of visual field. Good at detecting and analyzing movements. Can't see equiluminant colors. What (Ventral): object/Visual recognition. Plays role in judging significance of items it identifies.

72. Is sensitivity to spatial frequency dependent on the temporal frequency of the stimulus? What does this imply about fast motion on displays?

Yes - sensitivity to spatial frequency dependent on velocity of stimulus. [Ware has an example of this complex function.] - Up to 2 degrees/second, anything slower and you have full acuity and anything faster you have lower acuity. Specific combinations of spatial/temporal frequencies can trigger epileptic seizures

57. Is it possible to compress an image to below 1 bit/pixel? Explain. (Difficult question.)

Yes! If I am representing a grid of pixels, an 8 x 8-pixel grid will have 64 pixels (code w/ 64 different patterns; 2^6 = 64 bits). I am only going to use 64 patterns for 64 bits. Creating 6 bits for 64 pixels allows for around 0.1 bits/pixel. By creating a small # of patterns for total # of pixels, create something below 1 bit/pixel. In summary: when any algorithm does compression, it doesn't just consider what to do with each pixel, it considers the pattern; represent pattern with smaller # of patterns and can reduce considerably (e.g. edge only represents edges). ALSO... Predictability allows us to reduce number of possible patches to describe parts of image. #Patches = 2^x. X gives number of bits required. Patch size gives number of pixels. SO: 1024 10x10 patches gives 10/100 = 0.1 bits/pixel JPEG 1. Break the image into 8x8 blocks 2. Calculate the spectrum (DCT) 3. Quantize the spectrum as far as visually possible

54. Are natural scenes redundant? Describe 3 forms of statistical redundancy. Why is this important for understanding compression? Describe how the visual system takes advantage of this?

Yes. 1. Nearby points are highly correlated: similar amplitude spectra (center-surround operators are very efficient) 2. Data are sparse - events are localized in time and space; natural scenes have sparse structure not described by the correlations between pixels. When the correlations are removed, natural scenes are typically dominated by local sharp discontinuities (edges). The state-space of natural scenes is not Gaussian. 3. Data show continuity across space, time, and frequency: sparse structure shows continuity -The cross talk between neighboring neurons: neurons that are co-circular and co-axial are often connected and enhance connected contours (good continuation) Compression: Takes advantage of redundancy to make bets & more efficiently code Visual system: spatial resolution (respond only to local differences; periphery only long range differences); temporal resolution (respond only to local changes over time) Visual system takes advantage of this by using receptive fields that take differences in intensity rather than absolute values; neurons only fire when something changes in intensity

n/a

n/a