PowerPoint Poster Template

1
Contrast Matching With Complex and Natural
Stimuli of Varying Angular Size
Bridget C. Hendricks, James Comerford, Frank
Thorn, and Eli PeliThe New England College of
Optometry, Boston, MA
Results
Introduction
Discussion
Observers demonstrate contrast constancy for
suprathreshold natural stimuli despite defocus
and aberrations induced by viewing images which
subtend small angular retinal sizes. The amount
of constancy is consistent across images of
varying sizes, for all images that are able to be
resolved. However, contrast constancy seems to
fail for stimuli near threshold. The seeming
failure of contrast constancy for stimuli near
threshold was also found in our previous study in
which sine wave grating stimuli were used. From
previous studies we know that induced refractive
defocus and scatter blur have little effect on
contrast constancy, when tested using high
contrast gratings of varying spatial
frequency. These results suggest that there is
an adaptive mechanism responsible for maintaining
constancy of image perception across varying
spatial frequencies as we scan our natural
environment. Significant amounts of
compensation for blur have previously been shown
to occur rapidly, and exhibit almost complete
interocular transfer (Mon-Williams et al, 1998
Webster et al, 2002). It has been suggested that
the mechanism for blur compensation is a function
of gain control for spatial frequency-selective
channels within the visual cortex (Georgeson
Sullivan, 1975 Mon-Williams et al,
1998). Contrast constancy has important
implications for the study of perceptual
adaptation in low vision due to either optical or
neural blur.
Defocus and aberrations degrade the retinal image
of high spatial frequencies compared to low
spatial frequencies, resulting in decreased
contrast sensitivity at high spatial
frequencies. Georgeson and Sullivan (1975)
found that observers matched the contrast of high
and low spatial frequencies so that high contrast
images of all frequencies appeared subjectively
to have equally high contrast despite effects of
retinal blur. They call this effect contrast
constancy. Contrast constancy has obvious
implications for vision with uncorrected
refractive error and may also be important for
low vision patients (perhaps patients can have
the perception of sharp contours despite low
contrast optical and neural images). Webster
et al (2002) studied the effects of contrast
adaptation as it applies to stimuli such as
simple edges and grayscale images of natural
scenes. Their results show large and rapid
changes in the perception of image focus when
people view images with altered spatial
statistics. Our previous experiments have
explored the effects of refractive defocus and
scatter blur on contrast matching, demonstrating
contrast constancy with suprathreshold simple
grating stimuli. We further explore this effect
as it relates to more natural images by
evaluating effects of varying retinal image size
(i.e. shifting the spatial frequency spectra of
the image) on contrast matching of complex,
natural stimuli.



Image 1
Image 2
Both Images have a characteristic distribution of
spatial contrasts as captured by the 1/f
amplitude spectra typical of natural scenes.
Reduced images were generated using a Bicubic
interpolation algorithm for image resampling.
References
Method
Burton, G.J., Moorehead. I.R. Color and spatial
structure of natural scenes. Applied Optics. 26.
157-170, 1987. Georgeson, M.A. and Sullivan,
G.D. Contrast Constancy deblurring in human
vision by spatial frequency channels. J. Physiol,
252, 627-656, 1976. Mon-Williams, et al.
Improving vision Neural compensation for optical
defocus. Proc. R. Soc. Lond. B 265, 71-77,
1998. Webster, et al. Neural adjustments to
image blur. Nature Neuroscience. Sept. 5(9)
839-840, 2002.
Fourteen observers were presented with a
grayscale image of a face (image 1) and a forest
(image 2) at Michaelson contrasts of 0.075, 0.15,
0.30, and 0.60. The perceived contrast of these
images was matched with the same image subtending
different angular retinal sizes (2x, 1x, 0.5x,
and 0.25x standard size image). The 2x image
subtended a visual angle of 2.86 degrees the
1x image subtended a visual angle of 1.43
degrees the 1/2x image subtended a visual angle
of 0.76 degrees and the 1/4x image subtended a
visual angle of 0.38 degrees. The standard
image and the experimental image were presented
side by side on a 19 inch Samsung monitor, at a
distance of 3 meters. Subjects adjusted the
experimental image to match the standard image in
contrast by using the up/down arrows on a
keyboard. An ascending and descending method of
limits experimental paradigm was used.
Acknowledgments
Results
Funded by the New England College of Optometry,
and the following Grants NECO Vision Core Grant
5 R24 EY014817 NIH/NEI Grant 5 T35
EY007149 Special thanks to Image Specialist,
Hyunjeong (Jamie) Han, for generation of natural
image stimuli, and image analysis.
The averaged results of our 14 observers are
presented on the right. While contrast
sensitivity functions typically show a decrease
in sensitivity at high spatial frequencies
relative to lower spatial frequencies, contrast
matching functions were unaffected by the high
spatial frequencies represented in the ½ standard
and ¼ standard images, relative to the lower
spatial frequencies presented in the 1X and 2X
standard images, thus demonstrating contrast
constancy. Results for image 1 (face) and image
2 (trees) were similar. Results were similar for
all 14 observers. Subjects reported that the
1/4x size target, during the lowest contrast
condition (standard 0.075 Michaelson contrast),
was difficult to resolve. Stimuli that were
reportedly difficult to resolve may represent
sub-threshold stimuli. Contrast constancy
appears to fail as stimuli approach threshold.
Contrast matching performance for images
subtending visual angles of 2.6 degrees (2x
image), 1.43 degrees (1x image), 0.76 degrees
(1/2x image), and 0.38 degrees (1/4x image),
relative to a 1.43 degree (1x image). Contrast
matching functions were unaffected by the high
spatial frequencies represented in the 0.38
degree and 0.76 degree images relative to the
lower spatial frequencies represented in the 2.6
degree and 1.43 degree images. Contrast matching
functions for image 1 (face) and image 2 (forest)
are similar.