Imaging the Ghent Altarpiece: insights from an interdisciplinary research and restoration project

During the complex restoration process of the famous Ghent Altarpiece, some of its mysteries that puzzled art historians for a very long time were unraveled. Alongside the restoration, research projects were carried out that revealed so far unknown and surprising aspects of the painting. Two researchers who worked on this famous artwork share their insights and research methodologies.

The online Technical Art History Series: Digital Tools for Cultural Heritage, aims to stimulate scientific and interdisciplinary discussions on cutting edge research in computational imaging methods and sensing technologies, combined with art history and digital humanities, applied to Cultural Heritage. The series brings together researchers from the humanities and sciences, presenting their research to allow an open forum where advanced knowledge on some of the most intricate and fascinating problems in cultural heritage is shared, and opportunities for interdisciplinary collaborations are promoted. In related blog posts, we will reflect on each session. The series is organised by the Rijksmuseum, the Computational Imaging Department from the CWI, Amsterdam, together with the Venice Center for Digital and Public Humanities, University Ca’Foscari, Venice.

The second session of our Technical Art History series on Digital Tools for Cultural Heritage which took place on May 25th, 2021, revolved around the Ghent Altarpiece (1432), by the brothers Jan and Hubert Van Eyck. Since 2012, the renowned altarpiece has been undergoing conservation treatment carried out by the KIK-IRPA (the Royal Institute for Cultural Heritage in Belgium) in the Museum for Fine arts in Ghent, and is extensively researched with cutting edge diagnostic techniques.

Our invited speakers were Aleksandra Pižurica and Geert van der Snickt. Aleksandra is Professor in Statistical Image Modeling at Ghent University, where she also leads the research group Artificial Intelligence and Sparse Modelling. Her research is at the intersection of image processing, machine learning and information theory. Geert is a tenure track professor in the conservation-restoration department at the University of Antwerp. Both discuss their research contribution to Phase I (2012-2016), the treatment of the exterior panels, and Phase II (2016-2020), the lower panels of the interior. In the 1950s, conservators already discovered that parts of the altarpiece had been overpainted, however, they were lacking the technology and time to determine and characterize the overpaints exhaustively. During the current conservation treatment, new techniques were applied to investigate the condition of the painting.

Figure 1. Jan and Hubert Van Eyck, The Ghent Altarpiece, 1432. ©Sint-Baafskathedraal Gent – – Dominique Provost

Aleksandra recounts how during the investigation it became apparent that 70% of the surface of the outer panels was overpainted. When these overpaints were removed during the restoration process, a different style of painting became visible. The original work by Hubert and Jan Van Eyk is indeed more dramatic as can be seen, for example, in the elaborately painted draping of some of the garments. But, as Aleksandra says: ‘The most dramatic find when revealing the paint layers of the main panel was the expressive face of the lamb.’ (Figure 2) This more humanised face of the lamb attracted a lot of media attention when it was revealed.

Figure 2. Left: the face of the lamb before restoration. Right: the face of the lamb after the overpaint layers were removed. ©Aleksandra Pižurica

However, the original paint layers also contained more pronounced cracks and paint losses. ‘Restorers asked us whether we could develop an automatic method for [crack and] paint loss detection’, commented Aleksandra. Detecting and documenting such deteriorations is important for the decision making process for the actual restoration, but is a laborious manual task that is prone to errors: ‘The semi-automatic tools that are available also require a lot of manual work, and only allow for relatively rough annotation’, she confirms. Her work has focused on automating this task, using multiple imaging modalities of which some were recorded before the treatment (digital macrophotography, infrared macrophotography, X-radiography), and some while the treatment was in process (digital macrophotography and infrared reflectography), as shown in Figure 3.

Figure 3. The various imaging modalities used to investigate the Ghent Altarpiece. ©Aleksandra Pižurica

The alignment of such multi-modal data is a crucial step prior to any attempt to extract useful information (i.e., cracks and/or paint loss); however, this procedure was challenged by large data, imperfect alignment, as well as scarce and often erroneous manual annotations. Interestingly, crack patterns were used as landmarks to align these multiple images perfectly (Figure 4) by “relying on the fact that many of these [cracks] will reappear across different modalities”, she explains. From the registered images, Aleksandra and her team were then able to better detect cracks and repeat the registration procedure, thus creating an iterative approach which eventually gave a more reliable registration of those images.

Figure 4. Schematic overview of the alignment process of multiple imaging modalities by using crack patterns. ©Aleksandra Pižurica

For the paint loss detection her team used a set of image sections with the paint losses annotated manually by conservators (in particular, Bart Devolder), and then taught a machine learning algorithm to do the same. A machine learning model could then be used for the classification of fragments as paint losses or cracks. Once these were detected and labeled, automatic methods were developed to virtually fill in the gaps and cracks, thus obtaining a virtual restoration of the painting (Figure 5). As the physical paintings were also restored manually, this provided the opportunity for a qualitative comparison with the digital restoration.

Figure 5. Details from the panel “John The Evangelist”. Left: Detail of the painting before restoration. Middle: the paint loss detected by the algorithms. Right: Inpainting result on the detected cracks. ©Aleksandra Pižurica

Virtual restoration was for example used to make the text on a painted book more readable (Figure 6a-b) and to discover whether a real text was used or a combination of calligraphic characters. After a virtual inpainting of the cracks, the letters became more visible and hence suitable for further investigation.

Figure 6a. Left: Detail of the upper right panel on the outside of the altarpiece. Right: Magnified detail on the book. ©Aleksandra Pižurica

Fig. 6b. Left: Detail of the book showing the cracks in the paint. Right: Inpainting results on the detected cracks. ©Aleksandra Pižurica

The work by Aleksandra and her team is a great example of interdisciplinary research in which digital methods are used for a better understanding and preservation of cultural heritage. Given that the third restoration phase will start in 2022, she concludes with explaining how they will especially target the painting of pearls, present in the upper panels of the Ghent Altarpiece: “We will develop approaches able to extract some kind of digital signatures of the painted pearls with more recent machine learning methods.”

While the first talk focused on imaging the visible surface of the Ghent Altarpiece, the second one gave attention to chemical imaging, i.e. images containing information about the identification and location of the pigments present in the paint layers of the altarpiece. Geert van der Snickt explains how MA-XRF scanning (Macro x-ray fluorescence) was a key technique to characterize the overpaint during the conservation/restoration treatment of the altarpiece. MA-XRF scanning is a recently developed non-invasive imaging technique that maps the distribution of different chemical elements present at or just below the paint surface. The technique is relatively time-consuming: 60 days were needed to fully scan the verso side of the altarpiece (phase I), which measures circa 8m² of paint surface, and a total of 37 separate scans were collected. While the MA-XRF instrument was scanning the paint surface millimeter by millimeter, Geert jokingly mentions that there was luckily some time to grab a beer while waiting.

Figure 7. MA-XRF instrument (University of Antwerp) scanning the central panel of the Ghent altarpiece.
©Geert Van der Snickt

Geert highlights that the main challenges and goals for the MA-XRF technique were to visualize and confirm the extent of overpaint, and simultaneously assess the condition of the hidden underlying original Eyckian layers. The latter was a key aspect for the decision making process, as, for example, it would not make sense to remove the overpaint if the original layers were in very poor condition. MA-XRF, and in particular the distribution maps of heavier elements, proved to be very informative regarding the condition of the underlying layers, and allowed for a quantitative estimation of the percentage of still intact original paint.

Additionally, the pigment composition of the overpaint was also investigated and the presence of specific pigments pointed towards overpaint that predated the 18th century. Azurite, a copper-based blue pigment, was for instance found in the overpaint of the burgundy coloured mantle of Elisabeth Borluut, as is visible in the paint cross-section (Figure 8), where it is clearly separated from the original paint layers by an intermediate varnish layer. The blue pigment azurite fell out of use after the invention of Prussian blue around 1704. Therefore its presence in the overpaint made it possible to estimate the timeframe of when the panels were overpainted. Given that the extent of overpainting present was never suspected during earlier conservation campaigns before the 1950s, some scholars had attributed the visible brushwork for centuries to the van Eyck brothers, while in reality, the original Eyckian forms and rendering were hidden from sight.

Figure 8. Characterization of the build-up and overpaint in the burgundy colored mantle of Elisabeth
Borluut. ©Geert Van der Snickt

In the end, the MA-XRF scanning, as well as information from paint cross-sections taken from the altarpiece, provided objective chemical information that played an important role in defining the conservation strategy to remove the non-Eyckian overpaint. This had a significant impact on the conservation procedure. The decision to remove all overpaint added two additional years to the treatment of Phase 1 and consequently led to a considerable increase in the project budget. However, as Geert explained, the decision to remove the overpaint revealed the quality of the Van Eyck’s original brushwork, and the subtle and beautiful light modelling now visible are absolutely exquisite.

Figure 9. The macrophotograph taken during the overpaint removal of the group of female saints in the central
panel (the adoration of the lamb) shows the differences between the beautiful original brushwork of Van Eyck
(below) and the less sophisticated overpaint (above) ©ClosertoVanEyck

Geert also disclosed new findings of a recent scanning experiment of the Portrait of Margareta Van Eyck in the Groeningemuseum in Bruges, where some exciting changes were detected beneath the surface. This provides of course a cliff hanger for more to come.

Francien Bossema, Nouchka De Keyser, Dzemila Sero, Erma Hermens

A big thank you to Elisa Corrò from the Venice Center for Digital and Public Humanities, University Ca’Foscari, who organised the online session.

R. Sizyakin, B. Cornelis, L.Meeus, H. Dubois, M. Martens, V. Voronin, and A. Pižurica. Crack Detection in Paintings Using Convolutional Neural Networks. IEEE Access, 2020, DOI

A. Pižurica et al., Digital Image Processing of The Ghent Altarpiece: Supporting the painting’s study and conservation treatment’, IEEE Signal Processing Magazine, vol. 32, no. 4, pp. 112-122, July 2015, DOI

B. Cornelis, T. Ružić, E. Gezels, A. Dooms, A. Pižurica, L. Platiša, J. Cornelis, M. Martens, M. De Mey, I. Daubechies, ‘Crack detection and inpainting for virtual restoration of paintings: The case of the Ghent Altarpiece’, Signal Processing, Volume 93, Issue 3, 2013, Pages 605-619, ISSN 0165-1684, DOI

  • Geert is part of the AntweRp Cultural HEritage Sciences (ARCHES) research group. See also AXES.
  • Relevant publications by Geert’s team:

Van der Snickt, G., Dubois, H., Sanyova, J., Legrand, S., Coudray, A., Glaude, C., Postec, M., Van Espen, P., Janssens, K. (2017). ‘Large-Area Elemental Imaging Reveals Van Eyck’s Original Paint Layers on the Ghent Altarpiece (1432), Rescoping Its Conservation Treatment’, Angewandte Chemie, 129(17), 4875–4879. DOI

Van der Snickt, G., Dooley, K. A., Sanyova, J., Dubois, H., Delaney, J. K., Gifford, E. M., Legrand, S., Laquiere, N., Janssens, K. (2020). ‘Dual mode standoff imaging spectroscopy documents the painting process of the Lamb of God in the Ghent Altarpiece by J. and H. Van Eyck’, Science Advances, 6(31). DOI

We also invite you to have a look at the website Closer to Van Eyck where you can explore and wonder through the high resolution images taken before, during and after the restoration as well as several other analytical imaging techniques such as radiographs and infrared images.

Preview of the Third Session The third session will take place on Tuesday June 8th, from 16-17.30 (CEST) and will feature a presentation by Professor Holly Rushmeier (Yale University, New Haven, USA), titled ‘Tools for Making Sense of Cultural Heritage Data’.
Abstract: A challenge in cultural heritage documentation, analysis and communication is that relevant data is available in a wide variety of forms. Tools are needed to manage text, numerical output from instruments, images (both 2D and 3D), video and more. Further, a cultural heritage professional needs to observe all of this data to make informed decisions and communicate the basis for them. In this talk I will talk about tools for two scenarios – the data and analysis relevant to an individual object, and the data and analysis relevant to an entire site. I will provide examples of software we have developed and using the software to make sense of diverse data sets.

Please register here to receive the link and stay tuned:

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s