A statistical approach applied to Western landscape paintings reveals spatial composition patterns that could characterize different artists and style periods, according to a study. Works of art are typically described in qualitative terms rather than systematically analyzed with mathematical methods. Hawoong Jeong, Seung Kee Han, and colleagues used information theory to characterize the spatial composition of 14,912 digitally scanned Western paintings made over a 500-year period. An algorithm segmented paintings vertically and horizontally in sequential steps, from the most prominent to the least informative compositional features. Horizontal partitions outlined the sky, earth, and atmospheric color changes, whereas vertical partitions bordered trees, plants, buildings, and cliffs. The authors found that different compositional patterns characterized different artists, artistic styles, and time periods. Approximately 87% of the paintings were segmented horizontally in the first step, and the position of this dominant horizontal dissection shifted over time. Baroque paintings in the 17th century frequently featured dominant horizons below the midline of the painting, with the sky occupying a larger portion of the canvas. By contrast, the dominant horizons were located near the midline during the Rococo and Romantic periods, and near the upper one-third of the canvas during the Realism and Impressionism periods and in the 20th century. The approach could be applied to other art forms, such as photography, film typography, and architecture, to reveal macroscopic quantitative properties that may not be readily discernible, according to the authors.
###
Article #20-11927: "Dissecting landscape art history with information theory," by Byunghwee Lee, Min Kyung Seo, et al.
MEDIA CONTACTS: Hawoong Jeong, Korea Advanced Institute of Science and Technology, Daejeon, KOREA; tel: +82-42-350-2543; e-mail: <hjeong@kaist.edu>; Seung Kee Han, Chungbuk National University, Cheongju, KOREA; tel: +82-43-261-2274; e-mail: <skhan@chungbuk.ac.kr>
Journal
Proceedings of the National Academy of Sciences