Human sensitivity to face statistics computed on V1 similarity

In a biologically motivated recognition system we represent face images as convolution values with a set of multiscale and multiorientation Gabor wavelets (a simple model of V1). Based on their discriminative power on recognition tasks various face images were reconstructed from the Gabor wavelet representation to test whether the computed statistics had any psychophysical relevance. The result of our experiments indicate that human performance is sensitive to statistical information derived from our recognition system. The study provides evidence that higher face recognition areas, such as FFA in humans, potentially not only preserve, but also compute upon earlier V1 similarity space.