Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.604314
Title: Computations mediating global motion pereption
Author: Rocchi, Francesca
Awarding Body: University of Nottingham
Current Institution: University of Nottingham
Date of Award: 2013
Availability of Full Text:
Access through EThOS:
Abstract:
The visual system extracts global (overall) motion estimates from natural scenes to construct representations of the visual world. Any moving image may be characterised by different local regions with distinct directions (or speeds), that might be independent from each other. Each point of the visual field only yields a piecemeal representation of moving objects and so conveys limited (local) information of the visual environment. To reliably estimate image motion, the visual system combines ('pools') these local signals into a global percept. Yet, it is equivocal whether the computational steps leading to human perceptual decisions are guided by statistical regularities present within the visual environment or whether they are better explained in terms of structured decoding of neural population activity. The purpose of this thesis was to investigate which strategies the visual system employs to combine local vectors of motion into a global percept. This pooling process was explored by using a combination of psychophysical techniques and computational modelling. A common assumption is that the pooling of visual motion signals is synonymous with averaging. However, this might only lead to a simplistic explanation of how local signals are combined by the visual system. The experiments outlined in this thesis explored whether or not global motion perception always adopts a strict averaging strategy. As observers often reported perceiving simultaneously overlapping (transparent) surfaces, the relationship between perceived global motion and transparency perception was also investigated and characterised. The findings described in this thesis have shown that perceived global motion does not always coincide with an averaging strategy. They also suggest that the mechanisms underlying global direction and global speed perception might be different. The spatiotemporal pooling of local direction signals was generally well predicted by a neural mechanism-based computation. By contrast, a geometric average oflocal speeds (that gives more weight to low speeds) was a good predictor of global speed perception. The results of these experiments have also shown that transparency perception strongly affects perceived global speed. Although observers' ability to discriminate between overlapping motion patterns was modulated by stimulus-based characteristics, the balance between integration and segregation of local motion signals appeared to mainly rely on the presence of relatively low and high speeds within a moving scene. Finally, these findings suggest that the visual system might encode transparent motion by employing multiple independent velocity-tuned channels.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.604314  DOI: Not available
Share: