Purchase this article with an account.
Charles A. Collin, Byron O'Byrne, Luisa Wang; Effects of image background on spatial frequency thresholds for face recognition. Journal of Vision 2005;5(8):829. doi: 10.1167/5.8.829.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
A growing number of studies have investigated the question of which spatial frequencies, if any, are optimally useful for face recognition. To our knowledge, all of these studies have used face images with monochromatic backgrounds, usually medium-gray. A potential limitation of this methodology is that it does not accurately reflect the real-world situation to which results are to be generalized. That is, in the real world the visual system must recognize faces against a variety of backgrounds, and the spatial frequencies needed for face recognition may be different in these circumstances than when the background is homogenous. In this study, we investigated the differences in spatial frequency thresholds for face matching across three different types of backgrounds: 1) Monochromatic gray, 2) fractal noise, and 3) natural scenes. Observers were asked to find their matching threshold, using the method of adjustment, in a 4AFC match-to-sample task. That is, four faces were presented at the bottom of the screen, and a high-passed or low-passed face was presented in the middle of the screen. Observers were asked to adjust the cut-off of the spatial frequency filter to the point where they could just match the center face to one of the four comparison faces. Our results show small but consistent differences in threshold according to the type of background surrounding the face. Images with a fractal noise background elicited higher low-pass thresholds and lower high-pass thresholds than did the other two background types. There was no difference between monochromatic gray backgrounds and natural backgrounds. These data support the generalizability of results from studies using monochromatic gray backgrounds to real-world vision. However, the data also suggest that non-structured backgrounds can produce additional difficulty in recognizing spatially filtered face images. Additional data using the Method of Constant Stimuli are also being gathered.
This PDF is available to Subscribers Only