Style transfer for improved visualization of underdrawings and ghost paintings: An application to a work by Vincent van Gogh
Anthony Bourached, George H. Cann, Ryan-Rhys Griffiths, Jesper Eriksson and David G. Stork.
This project investigates how computational style transfer can improve the visualisation of paintings concealed beneath other artworks. Focusing on the two wrestlers hidden beneath Vincent van Gogh’s Still Life with Meadow Flowers and Roses, it applies colour and brushstroke characteristics learned from his surviving works to achromatic images of the underlying composition, generating a possible visual reconstruction.
The method combines neural-network-based edge detection, manual image preparation and GAN-based style transfer. The researchers extract contours from an edited X-ray image and label the figures’ skin regions. These guide the transfer of stylistic information learned from Van Gogh’s paintings, connecting the reconstruction’s colour and surface treatment to the composition revealed through imaging.
Extending the team’s earlier research, the project learns from distinct groups of artworks rather than treating Van Gogh’s style as uniform. Representative paintings from his Netherlands, Paris and southern France periods provide different palettes and brushstroke characteristics. These styles are learned separately and computationally combined before being transferred to the concealed composition. This approach explores how multiple stylistic references can contribute to a more coherent visualisation.
The study reports qualitative improvements in the alignment of colour with extracted contours, the consistency of lighting and shading, and brushstrokes that follow the figures’ forms. The result remains an interpretative reconstruction rather than a verified recovery of the original appearance. Facial rendering, uncertainty in contour extraction and validation against known images remain areas for further development. The project advances computational methods for examining hidden artworks while retaining the distinction between imaging evidence and inferred artistic style.
Image descriptions: Image 1 Reconstruction of Vincent Van Gogh, Two Wrestlers by Oxia Palus. Photo courtesy of Oxia Palus.
Image 2 Reconstruction of Titian, Psyche and Cupid by Oxia Palus. Photo courtesy of Oxia Palus.
In the News:
Related publications:
Style transfer for improved visualization of underdrawings and ghost paintings: An application to a work by Vincent van Gogh 2023
We applied computational style transfer, specifically coloration and brush stroke style, to achromatic images of a ghost painting beneath Vincent van Gogh’s Still life with meadow flowers and roses. Our method is an extension of our previous work in that it used representative artworks by the ghost painting’s author to train a Generalized Adversarial Network (GAN) for integrating styles learned from stylistically distinct groups of works. An effective amalgam of these learned styles is then transferred to the target achromatic work.
