Abstract
Abstract: :
Purpose: To determine to what extent the information content in a face image (i.e. number of pixels) can be reduced before compromising peripheral retinal face recognition performance. The results of this study might be used to identify the resolution required by electronic displays used in low vision devices suitable for people with central visual loss. Methods: Digitised photographs of six Caucasian faces were block averaged (pixelated) according to a logarithmic scale so that images contained 256, 1024, 4096, 16384 or 65536 pixels. All images were the same size (74mm2) and were presented sequentially at the centre of a monitor. The images were viewed eccentrically (7°) by 5 young and 5 elderly subjects. The presentation sequence was randomised both for the face presented and the level of pixelation. Prior to the experiment subjects were trained to associate a name with each face by watching a video. Subjects were required to identify each of the faces presented (6 AFC). By obtaining a series of measurements at different distances face recognition sensitivity, in terms of the visual angle subtended by the image, was determined. Results: Face recognition performance was only compromised when the information content of the images was significantly reduced and was comparable for images containing more than 16384 pixels per image. Conclusion: Our findings show that high resolution images (i.e. those containing many pixels) are unnecessary for the task of peripheral retinal face recognition. This suggests that high-resolution displays are not required for the task of face recognition in people with central visual loss.
Keywords: 411 face perception • 459 low vision • 310 aging: visual performance