Professor Hawoong Jeong from the Department of Physics has systematically analyzed the evolution of artistic styles and design paradigms in landscape painting using a computer algorithm. In his research, Professor Jeong collaborated with analysts, statisticians, and art historians from various institutions in Korea, Estonia, and the US.

One of the major goals in art history is identifying trends and changes in the principles of organization applied to paintings by artists, both purposely and subconsciously. However, due to art history research being predominantly conducted using qualitative methods, there has not been a large number of major quantitative and macroscopic analysis of these principles. In his research, Professor Jeong aimed to create a computational framework that assesses the fundamental organizational principle of compositional proportion in landscape paintings.

The study was based on a dataset of almost 15,000 landscape paintings created by more than 1,476 artists between the Western Renaissance and the 21^st century. The assessment framework created by the professor and his team employs Rigau’s dissection algorithm. The algorithm progressively subdivides an original painting using vertical and horizontal lines so that the resulting partitions proved maximum information gain. For example, if dissected subregions comprise a completely different color from other subregions, the information gain is maximized. On the contrary, if dissected subregions have the same color frequency distribution with the original image, the partitioning provides no meaningful information, and therefore, the information gain is minimized.

It turned out that some paintings provide more information when dissected horizontally rather vertically, and vice versa. The scientists divided the paintings into four groups according to their two most dominant dissections: horizontal-horizontal (H-H), horizontal-vertical (H-V), vertical-vertical (V-V), and vertical-horizontal (V-H). The proportion of H-H paintings in the dataset is the highest, followed by H-V and V-H type paintings, due to the fact that landscape images tend to have a distinct horizon line. The algorithm was able to identify the changes in the position of the horizon over time, which coincided with the chronology of emerging art movements. Interestingly, despite the high diversity in styles and topics in 20^th century art, paintings of that time share a similar placement of the horizon line at around one-third of the canvas height from the top.

However, Professor Jeong notes that “such a clean and systematic change in the composition of Western art history may reflect the actual history of art, but at the same time, we should keep in mind that [the results] may indicate the bias of the mainstream [consensus] in art history, which has been evaluated and organized by art historians and critics.”