Predicting Visual Complexity

Abstract

Inspired by the contrast between ‘classical’ and ‘expressive’ visual aesthetic design, this paper explores the ‘visual complexity’ of images. We wished to investigate whether the visual complexity of an image could be quantified so that it matched participants’ view of complexity. An empirical study was conducted to collect data on the human view of the complexity of a set of images. The results were then related to a set of computational metrics applied to these images, so as to identify which objective metrics best encapsulate the human subjective opinion. We conclude that the subjective notion of ‘complexity’ is consistent both to an individual and to a group, but that it does not easily relate to the most obvious computational metrics.