Abstract
The highest frequency disparity grating which can be perceived as a corrugated surface is around 3–4 cycles per degree (Tyler 1974; Bradshaw & Rogers 1999; Banks et al. 2004), much lower than the analogous limit for luminance gratings. This low limit has been explained in terms of an initial encoding of disparity into piecewise frontoparallel patches (Banks et al. 2004, Nienborg et al. 2004): the stereoresolution limit reflects the smallest windows across which interocular correlation is measured. Existing studies of stereoresolution have all used sinusoidal variations in disparity. This probably reflects practice in the luminance domain, where “over a wide range of spatial frequencies the contrast threshold of a grating is determined only by the amplitude of the fundamental Fourier component of its waveform (Campbell & Robson 1968). Yet there is no theoretical reason to expect this to hold in the disparity domain. Under the explanation proposed by Banks et al. (2004), higher stereoresolution would be expected with square-wave disparity gratings, in which disparity alternates between two values, than with sine-gratings in which disparity varies smoothly. Square-wave gratings are piecewise-frontoparallel, and so present less effective disparity “noise” to a matching process which assumes disparity is locally constant. Furthermore, recent evidence suggests that cells in V2 are specialized to detect disparity edges, which occur naturally at objects boundaries and are important in scene segmentation (von der Heydt et al. 2000, Bredfeldt & Cumming 2002). These edge-detectors would be expected to be driven more strongly by square-wave disparity gratings than by sines. We tested this expectation by comparing stereoresolution in random-dot patterns portraying horizontally-oriented disparity corrugations with square-wave and sinusoidal profiles. To our surprise, there was little difference in the stereoresolution measured with the two grating types. We discuss the implications for the piecewise-frontoparallel model of disparity encoding.