Purchase this article with an account.
Ian Scofield, Charles Chubb, George Sperling; Analyzing band-selective preattentive texture mechanisms. Journal of Vision 2008;8(6):352. https://doi.org/10.1167/8.6.352.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Purpose: To analyze preattentive mechanisms sensitive to variations in contrast in isotropic textures composed of narrow-band elements. Method: Textures were composed of small elements consisting of Difference of Gaussians (DoGs, aka Mexican Hats), all identical in spatial form but with eight different center contrasts varying from −1.0 (black center/light surround) to +1.0 (white center/dark surround). For all DoG elements, luminance averaged over the DoG was equal to the background luminance. Stimuli were scrambles (dense, spatially random arrays) of 2304 DoGs. In a 4AFC task, observers judged the location of a target scramble patch in a background scramble field. The difference d between contrast histograms of targets and backgrounds was experimentally varied. We assume that human vision has preattentive mechanisms differentially sensitive to different DoG-contrasts and that histogram differences dare discriminable only if the sensitivity function of at least one of these mechanisms has nonzero correlation with d. If we find a two-dimensional space of histograms in which some discriminations are possible but which also contains a maximum-amplitude difference dnull for which discrimination is at chance, then most likely only one mechanism is sensitive to differences in that space. Perturbation methods can then be used to measure the sensitivity of this mechanism to all 8 DoG contrasts. Results/Conclusions: We have been able to use this method to isolate and measure the sensitivity function of a previously unknown mechanism. This mechanism is highly sensitive to texture energy but also exhibits significant asymmetries in its sensitivity to positive vs. negative DoG contrasts. Human vision has other mechanisms sensitive to these DoG scrambles. For example, at least one additional mechanism sensitive to the sign of DoG contrast. By orthogonalizing the space of histograms to the sensitivity function of the mechanism we have found, we hope to isolate and characterize these other mechanisms.
This PDF is available to Subscribers Only