C. Redies and G. Hayn-Leichsenring
This project studies higher-order statistical properties of face images and their relation to the perception of individual characteristics of a person, with a particular focus on face attractiveness. The project’s aim is to determine how much information about the individual person characteristics can already be deduced from higher-order image statistics that are potentially processed at early stages of visual perception. The work is based on similar work on the statistical properties of natural scene images and aesthetic artworks, where the PIs have shown that the two types of images resemble each other in that both have a scaleinvariant (fractal-like) Fourier spectrum. Previously, many researchers have studied the Fourier spectral composition of face images in studies of face representation, typically with bandwidth frequency-filtered faces. These studies have demonstrated an effect of altering the spatial frequency profile of face images on face learning and recognition but revealed inconsistent results that favour different frequency ranges. Only a few studies have addressed the question how the spatial frequency spectrum and other statistical image properties affect the perception of face attractiveness.
Here we pursue three strategies to identify statistical image properties that may affect low-level perceptual mechanisms of face attractiveness and other individual person characteristics. First, we will correlate statistical image properties (Fourier spectral composition, features from pyramid histogram of gradient analysis, etc.) with person characteristics. Second, we will study the effect of these image statistics on the perception of individual characteristics in face images by psychological evaluation. Third, the neural underpinnings of the perception of statistical properties of face images will be probed in adaptation experiments and by recording ERPs from human participants. For comparison, aesthetic images (for example, face portraits by artists) and non-aesthetic images will be used to characterize the differences between face attractiveness (defined as the physical allurement of a person) and image beauty (defined as the pleasure derived from the global composition of an image). In summary, this project will provide a description of individual characteristics in face images in terms of their statistical properties as well as in relation to other categories of images. Moreover, we will study how a modification of these properties will affect the perception of individual characteristics in face images.
