The effect of spatial frequencies on perceived motion.
ABSTRACT
This paper investigates motion perception, and whether there are differences at different spatial
frequencies, to see if motion perception is impaired in blurred conditions. Graham (1979) suggested
there were multiple channels which were selectively sensitive to spatial frequencies. This paper
aimed to investigate motion, and replicate Campbell & Maffei’s (1981) findings that motion
perception is different at different spatial frequencies. A within-participants design was used, with
one participant. The observer was presented with sinusoidal gratings and had to respond whether
they had been displaced to the left or the right. A curve equation was fitted to the data to give
thresholds, and it fitted to 3 of the 5 spatial frequencies. The data showed that whilst there was
some evidence for different motion perception at different spatial frequencies, there was no clear
pattern.
INTRODUCTION
The question this paper aims to investigate is whether drivers are able to perceive motion accurately
when they have “blurred” vision of the road, i.e. in rain or fog. This has important practical
implications for driving safety.
In the early stages of the visual cortex, neurons have receptive fields which have a centre-surround
excitory -inhibitory set up. This means activation on the centre will excite the cell and activation on
the surround will inhibit the cell. (Mather 2006)Therefore a grating will activate a cell if the bars
match the receptive field. A grating with a low spatial frequency will activate neurons with large
receptive fields. Likewise, a grating of high spatial frequency will activate neurons with a small
receptive field, as the bars will fit in the centre of the field. This is a basic indication that high and
low spatial frequencies are responded to by different neurons.
Fourier analysis suggests that whatever we see in the world can be broken down into a basic set of
components. These components are known as sinusoidal gratings, where the “luminance in the
direction perpendicular to the stripes varies sinusoidally, while the luminance in the direction
parallel to the stripes is constant.” (Graham, 1979). Graham (1979) summarised that although there
is no evidence showing that the brain performs a strict Fourier analysis, there are numerous studies
suggesting the brain carries out a crude Fourier analysis; that it performs operations with many of
the basic traits of Fourier analysis.
, Gratings have four basic features; contrast, spatial frequency, orientation and phase. Contrast is the
magnitude of the intensity difference between the bars. Spatial frequency refers to how many bars
the grating contains per degree of visual angle. Orientation is the angle of the bars, and phase is the
bars position within the grating.
Hubel & Weisel (1962) studied area 17 cells in the cortexs of both cats and monkeys, and found cells
responded preferentially to stimulus with dark or light bars. They concluded simple cells responded
to certain bars, slits or edges in one position, and that these simple cells fed into complex cells. This
was known as the feature detector model and initially received a lot of experimental support.
However more recent knowledge has begun to question some of the key features of the model, and
whether the cells are actually feature detectors or just spatial frequency filters.
DeValois et al (1978, cited in DeValois & DeValois, 1980) showed that cells in area 17 were more
selective for spatial frequency than for bar width. Campbell & Robson (1968) provided further
evidence. They used sinusoidal gratings of different spatial frequencies to measure the lowest
contrast at which observers were still able to see the grating. Their data suggests the visual system is
breaking down the stimuli into individual spatial frequency components, supporting Graham’s (1979)
idea of a crude Fourier analysis.
Campbell and Robson’s research was pioneering in the field, as it first suggested there might be
multiple, narrowly tuned spatial frequency channels. DeValois & DeValois (1980) evaluated that
overall, evidence suggest processing in the striate cortex is more likely acting as spatial frequency
filters, not feature detectors.
Evidence for multiple spatial frequency channels is provided by experiments into masking (Wilson et
al 1983) and adaptation (Blakemore & Campbell 1969). The visibility of test patterns are measured
after the observer has observed a supra-threshold adaptation or a masking pattern. If the patterns
are processed by the same channel we can expect that visibility of the test pattern should be
affected by the adaptation or masking pattern. In fact, unless the patterns contain similar spatial
frequencies, the visibility of the test pattern is not affected. This provides further evidence that
channels are selectively sensitive to spatial frequencies. (Graham 1979)
An example of biological evidence into the differences in processing is provided by Tootell et al
(1998). They studied the striate cortex of macaque monkeys as the monkeys viewed sinusoidal
gratings of certain spatial frequencies. They found the pattern of 14-C-2-deoxy-d-glucose (DG)
produced differs depending on the spatial frequency. They also suggested that Magnocelluar cells
ABSTRACT
This paper investigates motion perception, and whether there are differences at different spatial
frequencies, to see if motion perception is impaired in blurred conditions. Graham (1979) suggested
there were multiple channels which were selectively sensitive to spatial frequencies. This paper
aimed to investigate motion, and replicate Campbell & Maffei’s (1981) findings that motion
perception is different at different spatial frequencies. A within-participants design was used, with
one participant. The observer was presented with sinusoidal gratings and had to respond whether
they had been displaced to the left or the right. A curve equation was fitted to the data to give
thresholds, and it fitted to 3 of the 5 spatial frequencies. The data showed that whilst there was
some evidence for different motion perception at different spatial frequencies, there was no clear
pattern.
INTRODUCTION
The question this paper aims to investigate is whether drivers are able to perceive motion accurately
when they have “blurred” vision of the road, i.e. in rain or fog. This has important practical
implications for driving safety.
In the early stages of the visual cortex, neurons have receptive fields which have a centre-surround
excitory -inhibitory set up. This means activation on the centre will excite the cell and activation on
the surround will inhibit the cell. (Mather 2006)Therefore a grating will activate a cell if the bars
match the receptive field. A grating with a low spatial frequency will activate neurons with large
receptive fields. Likewise, a grating of high spatial frequency will activate neurons with a small
receptive field, as the bars will fit in the centre of the field. This is a basic indication that high and
low spatial frequencies are responded to by different neurons.
Fourier analysis suggests that whatever we see in the world can be broken down into a basic set of
components. These components are known as sinusoidal gratings, where the “luminance in the
direction perpendicular to the stripes varies sinusoidally, while the luminance in the direction
parallel to the stripes is constant.” (Graham, 1979). Graham (1979) summarised that although there
is no evidence showing that the brain performs a strict Fourier analysis, there are numerous studies
suggesting the brain carries out a crude Fourier analysis; that it performs operations with many of
the basic traits of Fourier analysis.
, Gratings have four basic features; contrast, spatial frequency, orientation and phase. Contrast is the
magnitude of the intensity difference between the bars. Spatial frequency refers to how many bars
the grating contains per degree of visual angle. Orientation is the angle of the bars, and phase is the
bars position within the grating.
Hubel & Weisel (1962) studied area 17 cells in the cortexs of both cats and monkeys, and found cells
responded preferentially to stimulus with dark or light bars. They concluded simple cells responded
to certain bars, slits or edges in one position, and that these simple cells fed into complex cells. This
was known as the feature detector model and initially received a lot of experimental support.
However more recent knowledge has begun to question some of the key features of the model, and
whether the cells are actually feature detectors or just spatial frequency filters.
DeValois et al (1978, cited in DeValois & DeValois, 1980) showed that cells in area 17 were more
selective for spatial frequency than for bar width. Campbell & Robson (1968) provided further
evidence. They used sinusoidal gratings of different spatial frequencies to measure the lowest
contrast at which observers were still able to see the grating. Their data suggests the visual system is
breaking down the stimuli into individual spatial frequency components, supporting Graham’s (1979)
idea of a crude Fourier analysis.
Campbell and Robson’s research was pioneering in the field, as it first suggested there might be
multiple, narrowly tuned spatial frequency channels. DeValois & DeValois (1980) evaluated that
overall, evidence suggest processing in the striate cortex is more likely acting as spatial frequency
filters, not feature detectors.
Evidence for multiple spatial frequency channels is provided by experiments into masking (Wilson et
al 1983) and adaptation (Blakemore & Campbell 1969). The visibility of test patterns are measured
after the observer has observed a supra-threshold adaptation or a masking pattern. If the patterns
are processed by the same channel we can expect that visibility of the test pattern should be
affected by the adaptation or masking pattern. In fact, unless the patterns contain similar spatial
frequencies, the visibility of the test pattern is not affected. This provides further evidence that
channels are selectively sensitive to spatial frequencies. (Graham 1979)
An example of biological evidence into the differences in processing is provided by Tootell et al
(1998). They studied the striate cortex of macaque monkeys as the monkeys viewed sinusoidal
gratings of certain spatial frequencies. They found the pattern of 14-C-2-deoxy-d-glucose (DG)
produced differs depending on the spatial frequency. They also suggested that Magnocelluar cells
