Institute of Communications and Connected Systems


BBC Culture reports on investigative research led by EEE Researcher

11 November 2022

The work led by Prof Miguel Rodrigues, an academic member of the Institute of Communications and Connected Systems at the Department of Electronic & Electrical Engineering was featured in BBC Culture.

Miguel Rodrigues Goya article

The article, entitled 'The hidden images found in masterpieces' discusses how artificial intelligence techniques are being employed by researchers around the world to resurrect cultural artefacts, such as a painting.

ARTICT, or Art Through the ICT Lens, a research partnership between University College London, Imperial College London, and the National Gallery is one of the focal points of the article. Prof Miguel Rodrigues, leading the researchers at UCL, is also the Principal Investigator of the EPSRC-funded project.  

On their work being featured in BBC Culture, Prof Miguel Rodrigues comments:

“Recent advances in artificial intelligence technology offer an opportunity to address outstanding challenges in various domains. UCL EEE is working closely with museums to help reveal and reconstruct hidden or lost features in artwork.”

Their latest research output, ‘Mixed X-Ray Image Separation for Artworks With Concealed Designs’, published in the journal IEEE Transactions on Image Processing, reports on the use of a self-supervised deep learning-based image separation approach applied to X-ray images of Francisco de Goya's Doña Isabel de Porcel (circa 1805).

While X-ray radiographs have been a turning point in the field of art investigation, X-ray images of paintings with hidden sub-surface features are notoriously difficult for experts to interpret. This is because they include contributions from both the surface painting and the hidden designs. To help interpret mixed X-ray images better, the researchers have proposed a set of artificially intelligent algorithms, powered by deep learning, that would separate the mixed X-ray image into two hypothetical X-ray images. One of these reconstructed images is related to the X-ray image of the concealed painting, while the second image contains information only related to the X-ray image of the visible painting.

This new deep learning approach also enables art historians, conservators, and heritage scientists to better understand other images of paintings with mixed compositions such as infrared reflectograms, X-ray fluorescence scanning element distributions maps or outputs from reflectance imaging spectroscopy.

The restorative project on Van Eyck’s Ghent Altarpiece (1432) led by Dr Rodrigues at UCL, in collaboration with researchers from Duke University and the National Gallery, was also reported on in the article. The double-sided panels of the painting produce layered X-rays that are difficult for conservators studying each layer to interpret. Their paper, ‘Artificial Intelligence for Art Investigation: Meeting the Challenge of Separating X-ray Images of the Ghent Altarpiece’, demonstrated the use of artificially intelligent algorithms that separated the X-ray images into two distinct images. This helped experts study features from the front and back of the painting’s double-sided panels.

Deep learning approaches are now widely used to address challenges across most sectors, and the research led by Prof Rodrigues and his team is a crucial example of this.

The combination of traditional X-ray images with deep neural network algorithms can be used to solve critical problems present in art investigation today, one of them being conservation of irreplaceable works of art.


Image Credit

Francisco de Goya, Do na Isabel de Porcel (NG1473), before 1805. Oil on canvas. © The National Gallery, London. (a). RGB image. (b). X-ray image