Adaptive filtering in spatial vision: evidence from feature marking in plaids

Much evidence shows that early vision employs an array of spatial filters t uned for different spatial frequencies and orientations. We suggest that fo r moderately low spatial frequencies these preliminary filters are not trea ted independently, but are used to perform grouping and segmentation in the patchwise Fourier domain. For example, consider a stationary plaid made fr om two superimposed sinusoidal gratings of the same contrast and spatial fr equency oriented +/-45 degrees from vertical. Most of the energy in a wavel et-like (eg simple-cell) transform of this stimulus is in the oblique orien tations, but typically it looks like a compound structure containing blurre d vertical and horizontal edges. This checkerboard structure corresponds wi th the locations of zero crossings in the output of an isotropic (circular) filter, synthesised from the linear sum of a set of oriented basis-filters (Georgeson, 1992 Proceedings of the Royal Society of London, Series B 249 235-245). However, the addition of a third harmonic in square-wave phase ca uses almost complete perceptual segmentation of the plaid into two overlapp ing oblique gratings. Here we confirm this result psychophysically using a feature-marking technique, and argue that this perceptual segmentation cann ot be understood in terms of the zero crossings marked in the output of any static linear filter that is sensitive to all of the plaid's components. I f it is assumed that zero crossings or similar are an appropriate feature-p rimitive in human vision, our results require a flexible process that combi nes and segments early basis-filters according to prevailing image conditio ns. Thus, we suggest that combination and segmentation of spatial filters i n the patchwise Fourier domain underpins the perceptual segmentation observ ed in our experiments. Under this kind of image-processing scheme, registra tion across spatial scales occurs at the level of spatial filters, before f eatures are extracted. This contrasts with many previous schemes where feat ure correspondence is required between spatial edge-maps at different spati al scales.