Computational Neurobiology Lab, The Salk Institute, USA
Wednesday 22 November 2006, 16:00
Seminar Room B10 (Basement)
Alexandra House, 17 Queen Square, London, WC1N 3AR
Natural image statistics and spatial context in visual processing
Spatial context exerts a dramatic influence on neural processing and perception. The response properties of visual cortical neurons are substantially modulated by contextual surround stimuli. Perceptually, the presence of contextual information can give rise to intriguing misjudgements of local features, as in the widely documented tilt illusion. We motivate a neural model of contextual cortical processing from the statistical properties of natural images. We consider some bottom-up statistics, and formulation in terms of a generative model known as the Gaussian Scale Mixture. We also describe preliminary work aiming to link this form of model, through population coding, to the tilt illusion.