Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.565536
Title: The 'aperture problem' in complex moving scenes
Author: Kane, D.
Awarding Body: University College London (University of London)
Current Institution: University College London (University of London)
Date of Award: 2011
Availability of Full Text:
Access through EThOS:
Access through Institution:
Abstract:
The initial encoding of direction by mammals occurs in striate cortex by neurons with small receptive fields that are tuned to narrow bands of the spatiotemporal frequency spectrum. Individual neurons are unable to signal the global direction of 2D motion and are instead sensitive to the 1D component of motion perpendicular to a moving edge. To compute 2D velocity, it is necessary to integrate over a range of 1D velocity sensors. In this work I probe the ability of the visual system to compute 2D velocity from a range of stimulus classes, including naturally contoured scenes, natural scenes and a global-Gabor array. My research shows that the motion stream is highly sensitive to the distribution of local orientations present in a moving image, but is largely insensitive to their spatial second-order statistics. I present a computational model of two-dimensional motion processing that is able to derive precise estimates of 2D motion directly from complex natural scenes. The model produces errors when confronted with stimuli composed of anisotropic orientation configurations and is able to capture many of the biases and errors experienced by human observers. Finally, I argue that observers’ misperceptions of 2D motion does not reflect a sub-optimal 2D motion strategy, but reflects a compromise between the competing requirements of defining motions in a spatially discrete manner across space, and the ability to accurately estimate 1D motions, on which the computation of 2D velocity must rely.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.565536  DOI: Not available
Share: