¨Scale-invariant Statistical Properties of Aesthetic Art Images¨
June 22nd 2010 – Berlin School of Mind and Brain
Abstract
The Fourier power spectra of subsets of aesthetic art, other visually pleasing images and natural scenes share a specific statistical property. On average, they are scale invariant (fractal-like), which means that the relative prominence of coarse structure and fine detail remains constant for different image scaling. In contrast, real-world photographs of objects, faces, plants and scientific illustrations have Fourier power spectra that deviate from scale invariance. Moreover, art images possess more uniform scale invariance across image orientations than comparable real-world photographs. These results suggest that subsets of aesthetic images display luminance contrast statistics that are more pronounced than those of real-world image categories. The significance of these findings with regard to sensory coding in the visual system will be discussed.
–Christoph Redies is Professor and Director at the Institute of Anatomy, University of Jena School of Medicine. Together with Joachim Denzler from the Institute of Informatics, University of Jena, he studies statistical regularities in art images and other image categories. This research attempts to identify universal properties of art images that relate to the basic principles of sensory coding in the human visual system. In other research, Christoph Redies studies the molecular basis of neural circuit formation during vertebrate brain development. Besides his scientific work, he has a long-standing interest in the visual arts and pursues abstract painting as his hobby.
Curated by Daniel Marguiles