Upload
others
View
4
Download
0
Embed Size (px)
Citation preview
Why Use Sine Waves to Study Vision?
Linear systems analysis provides a compact mathematical description of visual performance.
Varying frequency allows us to measureperformance at all spatial or temporal scales.
Same description can be applied to the eye’s optics and to neural processing.
Can be applied in comparative studies of vision.
Fourier Analysis
FourierSynthesis
FourierAnalysis
Low Pass Filtered High Pass FilteredOriginal
The World Has Structure at Many Spatial Scales
Vernier Acuity5” of arc
1/6 the diameter of a foveal cone
Imax - Imin Imax + Imin
∆II
==Contrast
Imax∆I
IImin
Visual Angle, aa = tan-1 (s/d)
a
Sun and moon each subtend 0.5 degree of visual angle (30 minutes of arc)Index finger nail at arms length subtends 1 degree of visual angle1 deg = 291 microns on the retina
s
d
Demonstration of the shape of the contrast sensitivity function for luminance gratings. Spatialfrequency increases continuously from left to right, and contrast increases from top to bottom.
FIG. 5.1 Photopic luminance CSFs for human and macaque observers. Contrast sensitivity is thereciprocal of the contrast necessary to detect a pattern at threshold. Here it is plotted as a function oftest spatial frequency (from R.L. De Valois et al., 1974, Vision Res., 14, 75-81. Copyright 1974,Pergamon Journals, Inc. Reprinted by permission).
FIG. 5.2 CSFs for several different species. The data are all normalized relative to the point ofmaximum sensitivity for each function. Note the qualitative similarity in shape of the various functionsdespite their displacement along the log spatial frequency axis (data redrawn from Northmore &Dvorak, 1979 [goldfish]; Blake et al., 1974 [cat]; Jacobs, 1977 [owl monkey]; R.L. De Valois et al., 1974[macaque monkey]; Fox et al., 1976, and Fox & Lehmkuhle, personal communication [falcon]. VisionRes., Copyright 1974, 1977, 1979, Pergamon Journals, Inc. Reprinted by permission).
Temporal Contrast Sensitivity Function
Temporal Resolution Limit ~60Hz
FIG. 5.1 Photopic luminance CSFs for human and macaque observers. Contrast sensitivity is thereciprocal of the contrast necessary to detect a pattern at threshold. Here it is plotted as a function oftest spatial frequency (from R.L. De Valois et al., 1974, Vision Res., 14, 75-81. Copyright 1974,Pergamon Journals, Inc. Reprinted by permission).
Fourier Optics
Point Source
How good is this lens?
Object Plane Image Plane
Blurred Image
Impulse Response = Point Spread Function
A. Perfect Eye
Planar wavefrontSpherical wavefront
B. Aberrated Eye
Aberrated wavefront
Sources of Retinal Image Blur
DiffractionAberrationsLight Scatter
I(r) = [2J1(πr) /πr]2
Radius in radians, ro = 1.22λ/a
3 mm 440 nm 0.62'3 mm 700 nm 0.98'2 mm 555 nm 1.2'8 mm 555 nm 0.29’
FWHM ~= ro
PupilDia., a λ
Visual angle
1 mm 2 mm 3 mm 4 mm
5 mm 6 mm 7 mm
Courtesy A. Roorda
1 mm 2 mm 3 mm 4 mm
5 mm 6 mm 7 mm
Courtesy A. Roorda
Fourier Optics
Sine Wave Grating
How good is this lens?
Object Plane Image Plane
Blurred and Shifted Grating Image
Modulation Transfer Function
0
0.2
0.4
0.6
0.8
1
0 50 100 150 200 250
Mod
ulat
ion
Tran
sfer
Spatial Frequency (c/deg)
Diffraction8 mm pupil
632.8 nm
Eye's MTF3 mm pupil
Off Limits in Normal Viewing
Mod
ulat
ion
tran
sfer
Spatial frequency (c/deg)
0.01
0.1
1 2 mm 3 mm 4 mm
0.01
0.1
1
0 10 20 30 40 50 60
5 mm
0 10 20 30 40 50 60
6 mm
0 10 20 30 40 50 60
7 mm
Spatial Domain Frequency Domain
Object ObjectSpectrum
Generalized Pupil Function
Image Image Spectrum
Optical Transfer Function
Point Spread Function
Convolved w
ith
Autocorrelation
Equals
Equals
Fourier Transform
Inverse Fourier Transform
Multiplied by
Fourier Transform
Inverse Fourier Transform
Squa
red
mod
ulus
of
Four
ier T
rans
form
Courtesy of David Williams, University of Rochester
PointspreadFunction
RetinalImage
0.5 deg
Perfect eye(diffraction limited) MRB GYY MAK
5.7 mm pupil
ComputingRetinal Images with
ConvolutionObject
Double Pass ProcedureMeasuring the Eye’s MTF:
First PassSecond
Pass
Aerial Image
Point Source
Laser interferometry avoids blur from diffraction and aberrations
0.01
0.1
1
0 50 100 150 200 250
Mod
ulat
ion
Tran
sfer
Spatial Frequency (c/deg)
Eye's MTFsin incoherent light
8 mm pupil632.8 nm
2 3 4 5 6 7.3 mm
Eye's MTF with Interference Fringes
1
10
100
0 10 20 30 40 50 60
Cont
rast
sen
sitivi
ty
Spatial frequency (c/deg)
interferometry
normal viewing
CSFnv = MTF x CSFint
0
0.2
0.4
0.6
0.8
1
0 50 100 150 200 250
Mod
ulat
ion
Tran
sfer
Spatial Frequency (c/deg)
Diffraction8 mm pupil
632.8 nm
Eye's MTF3 mm pupil
Off Limits in Normal Viewing
High Frequency Cut: Partly Optical Blur
Ganglion Cell BodiesBeth Peterson and Dennis Dacey, UW
ON Center OFF Center
+-+
-
Center-Surround Antagonism in Ganglion Cell Receptive Fields
FIG. 3.16 Cross-sectional diagram of a bipolar cell receptive field profile. In the top drawing (A) thecenter and surround inputs are shown separately. Since they are opposite in sign and sumalgebraically, the resultant combined RF profile is as shown in B. In C is a three-dimensionalrepresentation of the entire RF, showing its radial symmetry (after the receptive field model ofRodieck, 1965).
FIG. 3.17 Superposition of a ganglion cell RF on gratings of different spatial frequencies. In themiddle figure the RF is superimposed on a grating whose light and dark bars are of such a width as tooptimally stimulate both RF center and surround. In the top figure the same RF is shown on a gratingof half optimal spatial frequency, and in the bottom, on one of twice optimal frequency. These latterpatterns would produce little net excitation or inhibition from the cell whose RF is shown.
Low Medium High
FIG. 3.16 Cross-sectional diagram of a bipolar cell receptive field profile. In the top drawing (A) thecenter and surround inputs are shown separately. Since they are opposite in sign and sumalgebraically, the resultant combined RF profile is as shown in B. In C is a three-dimensionalrepresentation of the entire RF, showing its radial symmetry (after the receptive field model ofRodieck, 1965).
Midget Ganglion Cell Receptive Field Center
+
Horizontal Cells
Midget Ganglion Cell Receptive Field
+ -
ON cells in layers 5 and 6
OFF cells in layers 3 and 4
I
I
I
C
C
C
Low Spatial Frequency:Little or No Response
Medium Spatial Frequency:Good Response
High Spatial Frequency:Little or No Response
Low Medium High
Lateral Inhibition? Finite Grating Summation Area Too!
0 20 40 60 80 100 120
104q
FT a
mpl
itude
(spi
kes/
sec)
0
10
20
30
40
50
60
70
20 40 60 80 100 120
94t
0
10
20
30
40
50
60
00
5
10
15
20
25
20 40 60 80 100 120 140
93y
0
Spatial Frequency (c/deg)
Monkey LGN Interferometry
Section Through Primary Visual CortexAlso called Striate Cortex or V1
FIG. 4.18 Three-dimensional representations of the RF profiles (in both space and frequency domains)of a fairly narrowly tuned cat simple cell. In the frequency domain plot it can be seen that the cellresponds to a delimited, compact range of spatial frequencies. Note in the space domain plot that thiscell has an oscillatory RF with multiple lobes along the x axis and is elongated in the y direction (fromWebster & R.L. De Valois, 1985, J. Opt. Soc. Am. A, 2, 1124-1132. Reprinted by permission).
FIG. 4.14 Quantitative simple cell receptive field profile (space domain) and corresponding spatialfrequency tuning function. The RF profile was measured by recording the responses (y axis) to anarrow, flickering black-white bar in different spatial positions (x axis). The solid lines in the columnon the left represent the RF profiles predicated by measuring the response to gratings of differentspatial frequencies (right). The data points are actual responses (from Albrecht, 1978. Reprinted bypermission).
FIG. 6.10 Macaque behavioral CSF (crosses) and the CSF of a single macaque striate cortex cell(circles). Note that the cell’s CSF has a very much narrower bandwidth than that of the overallbehavioral CSF. The latter presumably reflects the activity of many more narrowly tuned individualcells, as depicted in figure 6.1.
FIG 6.4 (A) CSF before (solid line) and after (data points) adaptation to a grating of a single spatialfrequency. (B) Threshold elevation at different spatial frequencies produced by the adaptation. Notethe band-pass loss in contrast sensitivity around the adaptation frequency (indicated by arrows) (fromBlakemore & Campbell, 1969. Reprinted b permission).
FIG. 6.7 Demonstration of the Blakemore-Sutton effect. First note that the two test gratings on theright are identical. Now adapt for about 1 min while scanning back and forth along the fixation barbetween the two gratings on the left, then quickly shift your gaze to the fixation line between the testgratings on the right. They should now appear different in spatial frequency, each one being shiftedaway from the frequency of the adaptation grating that occupied the same retinal area.
FIG. 6.8 A model of spatial frequency channels and the effects of adaptation, to account for theBlakemore-Sutton apparent spatial frequency shift. Three spatial frequency channels (A, B, and C)are presumed to signal the apparent spatial frequency in this region by their relative activity rates. Forinstance, x will have a certain apparent spatial frequency because it equally stimulates channels A andB. Following adaptation to a grating of the middle frequency, the sensitivity of channel B is reduced toB’. Now, a grating of the frequency y will equally stimulate channels A and B (now B’), and thus willlook the same as x had looked before adaptation.
FIG. 6.1 Spatial contrast sensitivity function as the envelope of many more narrowly tuned spatialfrequency selective channels.
FIG. 5.1 Photopic luminance CSFs for human and macaque observers. Contrast sensitivity is thereciprocal of the contrast necessary to detect a pattern at threshold. Here it is plotted as a function oftest spatial frequency (from R.L. De Valois et al., 1974, Vision Res., 14, 75-81. Copyright 1974,Pergamon Journals, Inc. Reprinted by permission).
Demonstration of the shape of the contrast sensitivity function for luminance gratings. Spatialfrequency increases continuously from left to right, and contrast increases from top to bottom.
Aqueous Humor
Vitreous Humor