BOTTOM-UP VISUAL IMAGE-PROCESSING PROBED WITH WEIGHTED HERMITE-POLYNOMIALS

This paper explores the use of Gaussian-weighted Hermite Polynomials ( WHPs) in studies of human spatial vision. WHPs form an orthonormal bas is set. They have several useful properties as visual stimuli, e.g., h aving a single origin which can be assigned to the fovea. Contrast-rev ersing WHP-modulated gratings of many levels of contrast, order, and s cale were used to elicit steady-state visual evoked potentials (VEPs) from human visual cortex. Pairs of WHP gratings were also superimposed and contrast-reversed, or presented separately and alternated, to obt ain sum and difference VEPs. We developed a model which related WHP de scriptors to VEP power with just three, model-based, parameters. The m odel avoids the use of numerous free parameters to describe essentiall y arbitrary modulation and phase transfer functions, and does not assu me the visual system is spatially homogeneous. The model provides a us eful ''black box'' account of gross cortical visual responses to patte rned stimuli.