CRS Guest Talks, Best Poster Prizes & Travel Awards > CRS Guest Lecturers > David Tolhurst
David Tolhurst is a senior lecturer in the Department of Physiology, Development and Neuroscience at Cambridge University. His research interests are in the area of the psychophysics and computational modelling of vision, primarily concerning the ways in which the visual system encodes the spatial, temporal and chromatic information found in the natural visual environment.
AVA Christmas Meeting 2006: Trying to Model Grating-Contrast Discrimination Dippers and Natural-Scene Discriminations
Campbell & Kulikowski (1966, J.Physiol., 187, 437-445) first described the contrast discrimination "dipper" for gratings. Since 1966, there has been increased quantification of dippers and increasingly sophisticated modelling that attempts to relate dipper form to the contrast responses of neurons in visual cortex and to nonlinear interactions between neurons. Understanding contrast discrimination is fundamental to models that attempt to explain the perceptual differences between complex 2D stimuli such as photographs of natural scenes (e.g. Lovell et al, 2006, ACM TAP, 3, 155-178). To develop such a model, we start with a model of grating detection performance which maps the behaviour of arrays of "simple cells" across the 2 dimensions of the stimuli and which includes untuned contrast normalisation (Watson & Solomon, 1997, JOSA A, 14, 2379-2391); we add convincing ideas about surround masking (Meese, 2004, J.Vis, 4, 930-943). Meese's surround model is needed to explain some unusual features in the form of dippers for different grating stimulus configurations, but seems initially incompatible with suggestions (e.g. Petrov et al, 2005, J.Neurosci, 25, 8704-8707) that there is no surround masking in foveal vision. We think that the incompatibility might be resolved with a full multi-channel 2D model (with several million stylised "simple cells") of the visual discriminations. A realistic spatial geometry of the receptive fields is crucial, along with the geometry of the suppressive surround. This would give the correct balance between the proposed surround masking and the counter effects of length summation between centre and surround stimuli.