Abstract
Optic flow and binocular disparity both describe changes in the visual direction of features when the optic centre translates (over time or from the left to right eye) and the eye rotates. However, different conventions have developed for describing this flow field, e.g. dividing it into 'horizontal' and 'vertical' disparity, which has no equivalent in the optic flow literature. Here, we ask whether there are principled grounds to use any particular 2-D coordinate system to describe retinal flow in general. We examine the special circumstances that apply to animals, in which the observer fixates a point as they move, and we explore the situations in which the epipole (e.g. direction of heading) is close to the fixation point or approximately 90 degrees away (as is commonly the case for binocular disparity). Any general formulation should include both. Flow in the direction of epipolar lines depends on local surface structure, whereas flow in the orthogonal direction does not. However, these directions are not fixed on the retina and so cannot be the basis of a 2-D retinal coordinate frame, despite the fact that binocular torsion goes a long way to keeping these directions constant at the fovea for different binocular eye positions. We argue that it can be helpful to divide retinal flow into radial flow (with respect to the fovea) and non-radial flow (e.g. Weinshall, 1990), because flow in these directions, respectively, carries information about (i) the component of optic centre translation towards the fixated point (whether this translation is the inter-ocular separation or observer motion) and (ii) the perpendicular component of translation. This division of retinal flow or disparity may be just as useful, we argue, as any division based on 'horizontal' and 'vertical' disparity and would avoid any discussion about a suitable definition of such terms.
Meeting abstract presented at VSS 2016