Scale - Space Recognition Based on the Retino - Cortical Transform

Visual transformations are often employed to achieve size and rotation invariances in computational approaches to vision. Nature appears to have adopted a similar strategy: research workers studying biological vision systems have known for over a decade that a particular spatial transformation appears to be applied to the visual signals that are conveyed between the retina and the primary visual cortex in mammals. This transformation, the Retino-Cortical (RC) Transform Wilson (1), has the property of re-mapping visual information to remove the effects of scale, 2D rotation and certain projective distortions, Johnston (2). In addition, this general class of transformations maps a dilating optical flowfield, i.e. a centred focus-ofexpansion, to a translating flowfield, Bartlett and Jain (3).