Spatial frequency tuning on the recognition of facial image

Abstract

This study examined the effect of soatial frequency on the recognition of facial identity and exoression. We measured the root-mean-square (RMS) contrast sensitivity and signal-to-noise ratio (S/N) for human observers in identity and exoression tasks. We simulated the ideal observer oerformance and calculated the efficiency from human and ideal observer's signal energy. These data suggest that the facial image is orocessed by a single Gaussian channel which is the most sensitive to around 20c/fw (cycle per face width).