1. Chapter 6
Spatial Vision

2. There is a rational order to the material we cover!
Ch. 1 – how to measure vision
Ch. 2 – absolute threshold (see anything?)
Ch. 3 – intensity discrimination (discriminate different intensities)
Ch. 4 – how do these thresholds change with adaptation to light and dark?
Ch. 5 – thresholds for seeing simple objects spatially arranged on the retina (lines, offsets, pairs of lines, etc.)
Now – Chapter 6, expand to measuring thresholds for detecting more complex objects (edges) and testing contrast sensitivity.

3. The visual system recognizes objects from patterns of light and dark.
We will focus on the psychophysical measurement of the mechanisms the visual system uses to process complex patterns of light and dark

4. A fundamental requirement for detecting objects is to determine the location of the edges (boundaries) of the objects

5. Absolute luminance is less important in spatial vision than relative luminance levels
…because object contrast is relatively constant
The reflectance from real objects does not vary (at least not very much), so
contrast = (LT – LB)/LB
remains fairly constant under different lighting conditions

6. The visual system responds to luminance differences (e. g
The visual system responds to luminance differences (e.g., contrast) more than to luminance
(the visual system is a very poor light meter), so
the brightness of an object is not always predicted by its luminance
We see this in three examples: brightness constancy, simultaneous contrast, & assimilation

7. Brightness Constancy
the brightness of objects is relatively invariant even though the absolute luminance varies widely.
Brightness is determined largely by relative local contrast,

8. Neural basis: against different backgrounds, the same increase in luminance produces the same response (V)

9. Simultaneous contrast : The brightness of an object is not always predicted by its luminance
The brightness of an object is affected by the local contrast with surrounding objects.
(brightness is increased when viewed against a background with which the object has positive contrast and reduced when the stimulus has negative contrast relative to the background)

11. Assimilation
the brightness of a stimulus co-varies with the brightness of a surrounding stimulus

13. On the left:
On the right:
Filtered to equalize luminance while accenting the contrast boundaries.
The visual system acts similarly to base brightness on local luminance changes
unaltered
darkened
lightened
Gradient
blurred

14. In review of what we’ve just covered:
A fundamental requirement for detecting objects is to determine the location of the edges (boundaries) of the objects
Absolute luminance is less important in spatial vision than relative luminance levels
…because object contrast is relatively constant
The reflectance from real objects does not vary (at least not very much), so
contrast remains fairly constant under different lighting conditions

17. Several examples: if you can’t see the boundaries (edges), you can’t see the object
“Good camouflage”

18. The visual system is so “wired” to see edges, it detect them when there is very little information to specify the edge.

20. To repeat: the visual system recognizes objects from patterns of light and dark
A fundamental requirement for detecting objects is to determine the location of the edges (boundaries) of the objects

21. Mach Bands are a brightness enhancement created by the visual system and demonstrate the accentuation of luminance changes at surface boundaries
Discovered by Ernst Mach (1865)
Mach also measured the speed of sound; “Mach One” is named in his honor

22. Mach bands - regions of increased or decreased brightness caused by the response of the visual system to luminance boundaries

23. Receptive field positions relative to light and dark panels
The responses of retinal neurons underlie both the dependence of brightness on contrast and the brightness enhancement of Mach Bands
Off On
Receptive field positions relative to light and dark panels 1 2 3 4
Neural responses

24. This is a really useful website for understanding how center-surround ganglion cells work. I will show it in class but you would benefit from viewing it before (or after) class.

25. Receptive field positions relative to light and dark panels
The responses of retinal neurons underlie both the dependence of brightness on contrast and the brightness enhancement of Mach Bands
Off On
Receptive field positions relative to light and dark panels 1 2 3 4
Neural responses

26. Center-surround organization
a more detailed look at how the surround interacts with the center
When light increases in both the center and surround, the surround subtracts from the center.

27. Light on in the receptive field center produces depolarization in the bipolar cell and increased firing rate (spikes/second) in the on-center ganglion cell
i = sign-inverting (metabotropic) synapses
c = sign-conserving (ionotropic) synapses

28. Light on in the receptive field surround produces hyperpolarization in the bipolar cell and decreased firing rate (spikes/second) in the on-center ganglion cell

29. The depolarization caused by light on in center is opposed by the hyperpolarization produced by light on in the surround

30. Thus, when both the whole center and the whole surround are stimulated with light, surround cancels almost all the depolarization caused by light in the center. The result is a very small depolarization and slight increase in ganglion cell firing (the center is slightly stronger than the surround)

31. Receptive field positions relative to light and dark panels
The responses of retinal neurons underlie both the dependence of brightness on contrast and the brightness enhancement of Mach Bands
Off On
Receptive field positions relative to light and dark panels 1 2 3 4
Neural responses
Receptive field center and surround of #1 are entirely in light. Surround subtracts from center, but center is a little stronger so cell responds a little to the high intensity

32. Receptive field positions relative to light and dark panels
The responses of retinal neurons underlie both the dependence of brightness on contrast and the brightness enhancement of Mach Bands
Off On
Receptive field positions relative to light and dark panels 1 2 3 4
Neural responses
Receptive field center and surround of #4 are entirely in darker band. Surround subtracts from center, but center is a little stronger so cell responds a little to the low intensity, but less than #1

33. Receptive field positions relative to light and dark panels
The responses of retinal neurons underlie both the dependence of brightness on contrast and the brightness enhancement of Mach Bands
Off On
Receptive field positions relative to light and dark panels 1 2 3 4
Neural responses
Receptive field center of #2 is are entirely in light. Only part of the surround is in the light, so the surround subtracts less from the center, so cell responds more strongly than #1. This causes the brightness increase at the right edge of the light bands.

34. Receptive field positions relative to light and dark panels
The responses of retinal neurons underlie both the dependence of brightness on contrast and the brightness enhancement of Mach Bands
Off On
Receptive field positions relative to light and dark panels 1 2 3 4
Neural responses
Receptive field center of #3 is entirely in the darker region, so the center is not strongly driving the cell. Some of the surround is in the light, so it subtracts strongly from center, so cell responds less than #4, causing the brightness decrease at the left edge of the dark bands.

35. Another reason having a receptive-field surround is important!

36. The sensitivity (of the visual system) to spatial luminance changes is measured with gratings of varying spatial frequency and contrast

37. Sensitivity of spatial luminance changes is measured with gratings of varying frequency and contrast

38. Sine-wave gratings are measured in terms of their spatial frequency defined as the number of cycles per degree of visual angle

39. Michelson (grating) CONTRAST
When periodic stimuli like gratings are used, contrast is
defined
as:
Contrast L = - + ( ) /
max
min
where L
max
is the maximum luminance in the grating, and L
min
is the minimum
luminance in the grating.
In a sine-wave grating, L
max
and L
min
are symetrically arranged around the
average
luminance, which is
defined
as: (L
max
+ L
min
)/2
where L
is the maximum luminance in the grating, and L
is the minimum
max
min
luminance in the grating .

40. The spatial CSF is the inverse of the contrast threshold to detect various spatial frequencies

41. The spatial CSF is the inverse of the contrast threshold to detect various spatial frequencies

42. The spatial CSF is “band pass
The spatial CSF is “band pass.” We are more sensitive to a “band” of mid-range spatial frequencies (3 – 10 cycles/deg)
Peak contrast
1000
sensitivity
Low spatial
frequency rolloff
100
Contrast Sensitivity
Snellen 6/6
Equivalent 10 1
0.1 1 10
100
Spatial Frequency (cycles/deg)
Cutoff high
spatial frequency

43. As shown in Chapter 5, the cutoff high spatial frequency (the highest spatial frequency that can be resolved at a contrast of 1.0) is very close (nearly identical) to the spatial resolution acuity (MAR)

44. How to measure the spatial CSF
Pick a series of spatial frequencies (cycles per degree)
At each spatial frequency, determine the threshold contrast
Contrast sensitivity is the inverse (reciprocal) of the threshold contrast
Plot the contrast sensitivity for that all the spatial frequencies that were measured
Connect the data points with straight lines

45. Spatial Contrast Sensitivity Function
(Spatial CSF)

47. The band-pass shape of the spatial CSF is the same in many species (but peak at different spatial frequencies)

48. Optical defocus (blur – uncorrected refractive error) affects high spatial frequencies more than low
The cutoff high spatial frequency matches the optical limit

49. So far in this chapter:
1) Detecting the boundaries of objects is essential for seeing objects
2) The center-surround receptive fields of retinal ganglion cells cause the visual system to emphasize luminance changes (contrast) that typically occur at the edges of objects (this produces Mach bands)
3) The sensitivity of the visual system to spatial changes is measured by the spatial contrast sensitivity function (CSF)
Now: the individual neurons (retina, LGN, visual cortex) have a band pass spatial CSF with more narrow tuning than the whole CSF due to receptive field size

50. Textbook figure Original figure

51. Underlying the spatial CSF are individual neurons, each with a spatial CSF
High contrast sensitivity
Low (0) contrast sensitivity
High contrast sensitivity
What is plotted is relative amplitude (strength, firing rate, # of action potentials/sec) of neural response to a constant contrast.
This is based on the fact that neurons respond more strongly to a stimulus that is farther above threshold
Low (0) contrast sensitivity
High contrast sensitivity
Low (0) contrast sensitivity

52. A retinal cell’s center-surround receptive-field organization produces a spatial CSF for that cell
In previous slides, we saw how the surround subtracts from the center (when light increases in both the center and surround)
Now, we see how the surround adds to the center (when light increases in the center and decreases in the surround)

53. Sensitivity profile of the
Receptive-field center and
surround
Light in center & light in
Surround; center slightly
Stronger. Will respond but
contrast threshold is high
(low sensitivity)
Good fit; strong response;
Therefore, lowest threshold
Light & dark in center cancel each
other; light and dark in surround
Cancel each other;
infinite threshold (cut-off)

57. The surround adds to the center (when light increases in the center and decreases in the surround)

59. So far in this chapter:
1) Detecting the boundaries of objects is essential for seeing objects
2) The center-surround receptive fields of retinal ganglion cells cause the visual system to emphasize luminance changes (contrast) that typically occur at the edges of objects (this produces Mach bands)
3) The sensitivity of the visual system to spatial changes is measured by the spatial contrast sensitivity function (CSF)
4) the individual neurons have a band pass spatial CSF with more narrow tuning than the whole CSF due to receptive field size

60. NOW:
Cells with a variety of receptive-field sizes provide the basis for the behaviorally measured spatial CSF
Test the whole creature’s threshold, rather than the threshold for an individual neuron

61. Do neurons with different receptive-field sizes produce the human spatial CSF?

62. Contrast adaptation demonstrates that there are multiple spatial frequency channels in the human visual system
Blakemore and Campbell (1969); contrast adaptation with 7.1 cycles/deg

66. This was a test of contrast adaptation at a particular spatial frequency.
Blakemore and Campbell measured contrast adaptation using other spatial frequencies as well

68. Magnocellular pathway neurons
Parvocellular pathway neurons

69. So far in this chapter:
1) Detecting the boundaries of objects is essential for seeing objects
2) The center-surround receptive fields of retinal ganglion cells cause the visual system to emphasize luminance changes (contrast) that typically occur at the edges of objects (this produces Mach bands)
3) The sensitivity of the visual system to spatial changes is measured by the spatial contrast sensitivity function (CSF)
4) the individual neurons have a band pass spatial CSF with more narrow tuning than the whole CSF due to receptive field size
5) the spatial CSF is produced by neurons with more narrow tuning than the whole CSF (e.g., “channels”)

70. The spatial CSF varies with luminance
As mean luminance decreases, peak contrast sensitivity decreases, the spatial frequency of the peak decreases, cutoff (acuity) decreases and low frequency rolloff becomes less

71. The spatial CSF shifts toward lower spatial frequencies with increasing retinal eccentricity

72. Applying the idea of spatial frequency channels to looking at real objects in the world:

73. We are not always presented with high-contrast images:

74. Applying the idea of spatial frequency channels to looking at real objects in the world:
All spatial luminance patterns on the retina (from real objects in the world) are comprised of sine-wave gratings of particular spatial frequencies and contrasts
Different subsets of neurons respond to the pattern of spatial frequencies present in the image

75. All spatial luminance patterns are comprised of sine-wave gratings of particular spatial frequencies and contrasts

76. Luminance profile of a single horizontal line in the photo

77. Fourier Analysis: can determine the component sine waves & contrasts of complex waveforms

79. Do a Fourier analysis of this single line to get the component spatial frequencies and contrasts

80. This shows the spatial frequencies that are present in the image along that narrow line. There are many different spatial frequencies present, and they occur in differing amounts (relative contrast)

81. The high frequencies are not detected by our visual system and the low spatial frequencies are attenuated by our visual system, emphasizing the mid-range spatial frequencies

82. Fourier Synthesis: can construct complex waveforms by adding together simple ones
Can use this to predict what we will see!

83. Fourier Analysis

84. Obvious examples of spatial frequencies in the world

85. O.K. The visual scene contains many spatial frequencies.
Do we have to have high spatial frequencies (near the acuity limit) to “see objects”, or can we “see objects” with just low & intermediate spatial frequencies?

86. Rev 2

87. We CAN “see objects” with just low & intermediate spatial frequencies!

88. Good contrast sensitivity is more important than resolution for some spatial tasks
We do not always need to see in high contrast conditions

89. VA measures “quantity” of vision
spatial CSF measures “quality” of vision

90. In the Pelli-Robson chart, letter size remains the same, but contrast decreases

91. Mobility: Pelli-Robson chart performance is a good predictor of time for patients with age-related macular degeneration to complete an obstacle course & number or errors (collisions).
Low vision patients: reading speed is affected by spatial CSF losses
Possible new treatments: selective contrast enhancement at certain spatial frequencies to boost face recognition

93. Many forms of ocular pathology affect the spatial CSF
glaucoma, ARMD, diabetic retinopathy, cataract, keratoconus
but many of these affect high spatial frequencies and are detected by loss of VA
no unique patterns of CSF loss have emerged that are of use in differential diagnosis of a particular condition
VA is easier to measure and almost as effective for most conditions.
But, at some time, you will see a patient whose VA is OK but they complain about difficulty seeing, and have a spatial CSF deficit.