Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.605738
Title: Vision as informed by the statistics of the natural environment
Author: Josephs, Jennifer Anne Elizabeth
Awarding Body: University of Southampton
Current Institution: University of Southampton
Date of Award: 2013
Availability of Full Text:
Access through EThOS:
Full text unavailable from EThOS. Please try the link below.
Access through Institution:
Abstract:
The natural statistics approach is based on the theory that there is a strong relationship between perceptual systems and the statistical properties of the environment they evolved in. The experiments contained within this thesis were varied in focus and methodology, but were all centred on characterising this relationship. Psychophysical observations suggest that human observers can exploit regularities and differences in amplitude spectra to perform 'high level' tasks such as scene categorisation. In two carefully controlled priming experiments, I found no evidence to support this idea. In another experiment, I investigated whether observers could use the sign or magnitude of 2D contour curvature to influence estimates of metric depth near the boundary. Evidence demonstrated that participants were significantly influenced by the magnitude of contour curvature, suggesting that observers can use the learned relationship between 2D and 3D surface properties to recover depth when other cues are unreliable. Lastly, I developed a novel stimulus and response set-up that allowed recording of an observer's perception of slant about multiple tilt axes. Using this novel technique, it was demonstrated that observers demonstrate a frontoparallel prior. This work suggests that measurement of the statistical regularities in natural scenes has the potential to explain many aspects of visual performance. For future work, a complete characterisation of these relationships will require systematic measurements of the properties of the natural environment and combining research knowledge across carefully controlled, novel psychophysical tasks.
Supervisor: Graf, Erich Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.605738  DOI: Not available
Keywords: BF Psychology
Share: