Paintings are often made up of thousands of tiny brushstrokes, each going in a certain direction, that are not easily observed by the viewer. A cross-disciplinary research team from the Penn State College of Information Sciences and Technology (IST) and Loughborough University in England has developed an image analysis method that helps to make the underlying brushstroke structure of paintings visible, giving new insight into how artists physically created their works.
This approach offers both experts and non-experts a fresh way to observe and interpret the making of artworks. The research was recently published in the journal Patterns.
The researchers bridged art and data science to show that painting style can be quantified and visualized as flow, turning elusive qualities like "gesture" into measurable, analyzable data. They used a computational technique to examine very small patches of Impressionist paintings, determining the direction of the brushstroke in each tiny spot and connect these different directions, as if drawing lines that follow the flow. This resulted in a set of "streamlines" that trace how the artist's hand and brush moved across the canvas. The study also measured features of the brushstroke flows - length, curvature, direction - so that different artists' styles could be compared.
"This work demonstrates how computer vision and data science can reveal subtle structural patterns in paintings that are difficult for the human eye to detect directly," said co-corresponding author James Wang, distinguished professor in the College of IST's Department of Informatics and Intelligent Systems. "Our method transforms hidden brushstroke information into a visual representation that supports deeper analysis of artistic technique and style."
The streamline visualizations offer a new lens for viewing and interpreting art, according to co-corresponding author Kathryn Brown, reader in art history and digital heritage at Loughborough University.
"They help observers - whether experts or general viewers - better understand how the artist moved their brush, how the painting is organized and how artists' styles differ," Brown said. "Essentially, we have a new computational 'roadmap' for interpreting the development of a painting."
In addition to Wang and Brown, contributors to this research included Lizhen Zhu, a graduating doctoral candidate in informatics advised by Wang, and Chaewan Chun, also a Penn State doctoral candidate in informatics.
The Penn State researchers were supported in part by the U.S. National Science Foundation.
