The aim of experimental (neuro-)aesthetics, an emerging field of neuroscience, is to elucidate the biological basis of esthetic perception. The central idea of neuroaesthetics is that artists create their artworks so that it reflects basic functional properties of the human visual system. For a theoretical perspective, see Redies (2007), and for a recent review, see Graham and Redies (2010).
Together with the group of Joachim Denzler, we discovered that graphic art from the Western hemisphere and natural scenes share self-similar (scale-invariant) statistical properties in the Fourier domain (Redies, Hasenstein and Denzler, 2007). A particularly striking example are human faces. Artists depict human faces in their artistic portraits with the statistics of complex natural scenes, although photographs of human faces are not scale-invariant (Figure 1; Redies, Hänisch, Blickhan and Denzler, 2007).
Figure 1. Log-log plots of radially averaged Fourier power versus spatial frequency. Average curves are shown for datasets of passport-type face photographs, face portraits by artists and complex natural scenes. Note that the curves for art portraits and natural scenes have a more shallow slope (about -2) than face photographs (about -3.5). From: Redies et al. (2007b)
If photographs of human faces are modified so that they display the same Fourier statistics as artistic portraits, the modified face images elicit event-related potentials (ERPs) in the human brain that indicate a facilitation of face recognition and learning (Figure 2; Blickhan et al., 2011).
Figure 2. 1/fp Characteristics of the Fourier power spectrum affects ERP correlates of face learning and recognition. ERPs show larger N170 and N250 amplitudes (arrows) for images with enhanced fine structures (red line), and smaller amplitudes for images with enhanced coarse structures (blue line), compared to unmanipulated images (black line). From: Blickhan et al., 2011)
In more recent work, we used the Pyramid of Histograms of Orientation Gradients (PHOG) method to confirm that aesthetic artworks of Western provenance are highly self-similar at different scales of spatial resolution (Amirshahi et al., 2012). Measuring self-similarity and other aesthetic features derived in the PHOG analysis allows us to distinguish the subset of colored artworks analyzed by us from 11 other image categories of natural and man-made scenes, patterns and objects (work in progress).
In future work, we aim to identify other statistical image properties that chararcterize aesthetic artworks, by using modern approaches in computer vision. In psychological experiments, we will also investigate whether any of the statistical properties identified by us are necessary and/or sufficient to induce aesthetic perception. Using electrophysiological methods, we will study how changing statistical properties affects human brain responses to aesthetic and non-aesthetic visual stimuli.
Supported by DFG (RE 616/7-1, funding period 2012-15).
Previous Work in Vision Research
Together with Lothar Spillmann, I studied the neon color effect in the Ehrenstein figure and other displays (Redies and Spillmann, 1981; Redies et al., 1984). Here is a display showing this visual illusion as well as modifications that weaken or
destroy it (Figure 3).Figure 3. Colored crosses inserted into the gaps of the Ehrenstein figure (a) induce a colored "neon"-like glow around them (b). In addition, the inserted crosses appear less contrasted to the background than te same crosses presented in isolation (c). Rotation or lateral displacement of the colored lines destroys the illusion (d). A circle placed around the cross also abolishes the neon-like glow (e). For a brightness version of the neon effect, see (f).
In other work, I demonstrated, for the first time, that neurons in the primary visual cortex (area 17) respond to subjective contours:
From: Redies, Crook and Creutzfeldt (1986) Exp. Brain Res. 61:469-481.