Visual pattern encoding with weighted hermite polynomials.

The human visual system is spatially inhomogenous, and this property should be included in models of visual processing. We used weighted Hermite polynomials (WHPs) to encode and to characterize such inhomogenous processing. Simulations using an order-transfer-function (OTF) defined for each WHP order, at three spatial scales, provide elegant predictions of spatial frequency discrimination, of WHP order discrimination, and variations in two-point resolution and detection sensitivity with retinal eccentricity.