4174
Abstract: Computational aesthetics is emerging as a new interdisciplinary field bridging computer science, arts, philosophy, and cognitive science. It investigates how modern technology can assist art in enhancing our understanding of aesthetic perception, evaluation, and meaning. Although widely used in various fields, including fine art, photography, calligraphy, and graphic design, limited research has been done on the aesthetic assessment of paintings. Existing literature attempts to formalize aesthetic measure into mathematical formulas as combinations of complexity and order, and defines perceptual features influencing aesthetic judgment. Based on these theories on aesthetic measure, we attempt to map the abstract mathematical forms and formulas in art theory to real computation in paintings and seek a way to analyze aesthetic aspects of paintings. For the feasibility of our research, we narrow variances in paintings’ styles and contents by focusing on specified types of paintings, i.e., Chinese ink paintings and Western oil paintings.
As one of two important components in the aesthetic measure, complexity is firstly investigated in our research. We evaluate visual complexity of Chinese ink paintings by exploring how three main components of a painting, i.e., chromatic space, white space, and stroke, influence the paintings’ perceived complexity. Two studies are conducted in this research. In the first study, we gauge three visual features of the paintings as independent variables and establish regression models to predict the perceived complexity of the paintings and conduct a robustness check of our results by including additional control variables. In the second study, we perform an eye-tracking experiment to collect empirical evidence to validate our findings based on subjectively reported visual complexity.
We further study the visual order of Chinese ink paintings by proposing measurements for several features closely related to visual order. People’s aesthetic rating scores on the visual order are used to represent the perceived order of the paintings. Via regression model, the center of gravity, symmetry, and white space are proved to affect the visual order of the paintings.
The cultural difference on different types of paintings is also discussed. We conduct a comparative study of visual order between Western oil paintings and Chinese ink paintings, concentrating on two dimensions of visual order, color, and composition. By comparing statistical distributions of extracted features, we analyze the difference and similarity between the two types of paintings in a computational perspective. A classification technique is used to find features that best distinguish the two types of paintings. We also discuss the possible relationships between the feature differences and the cultural backgrounds.
Department: , The University of Texas at Dallas
Advisor(s): Kang Zhang
