Frequency Magnitude and Images of Texture Studies on Relationships to Human Preference

Abstract

A relationship between spatial frequency magnitude and aesthetic preference for texture is established in this work. This paper examines the mean output of three frequency settings in terms of preference ratings for a bank of Gabor filters. Three studies were conducted and the correlations between aesthetic ratings and the mean output for the filter set to extract frequency content at the scale of 0.35 cycles/ pixel were robust. The correlations for the mean frequency magnitude of this filter remained significant when image identifiability was incorporated; suggesting that memory and association are not exclusively driving (aesthetic) preference. These results are consistent with findings reported by Albrecht and Geisler (1997), demonstrating that human beings are tuned to specific frequencies and orientations. Overall, the results reported in this paper begin to substantiate a claim that specific frequencies contained in images do play a significant role in human preference. 1. Goodness of Form Prägnanz Early in the last century the quality of a visual experience was approached by Gestalt psychologists through the analysis of ‘goodness’ of form, or prägnanz, whichin their opinion was an important factor in perceptual experience (Beardslee et al., 1923; Kofka, 1922; Mowatt, 1940; Kohler, 1947). Perceptual “experience” was said to correspond to global properties of subjective simplicity, complexity and bilateral symmetry. Researchers did not believe whole figures could be understood by means of examining the local properties of those figures. Some images were thought to produce experiences of greater simplicity, order and regularity than others. Stimuli that are simple (contain adequate redundancy), symmetric (rotationally invariant) and familiar were considered “good”. However, a reliable description for judged simplicity was not determined and the relationship between symmetry and figural “goodness” (prägnanz) was not demonstrated. Symmetry proved only to be a tenable explanation and local asymmetry was not addressed (Garner 1962, 1970).