Allure of Craquelure: A Variational-Generative Approach to Crack Detection in Paintings

Researchers proposed a hybrid variational-generative approach for detecting craquelure in digitized paintings, modeling crack detection as an inverse problem that decomposes an image into a crack-free painting and a crack component. The method uses a deep generative model as a prior for the artwork and a Mumford-Shah-type variational functional with a crack prior to produce pixel-level crack maps, aiding art conservation.

arXiv:2602.09730v3 Announce Type: replace-cross Abstract: Recent advances in imaging technologies, deep learning and numerical performance have enabled non-invasive detailed analysis of artworks, supporting their documentation and conservation. In particular, automated detection of craquelure in digitized paintings is crucial for assessing degradation and guiding restoration, yet remains challenging due to the possibly complex scenery and the visual similarity between cracks and crack-like artistic features such as brush strokes or hair. We propose a hybrid approach that models crack detection as an inverse problem, decomposing an observed image into a crack-free painting and a crack component. A deep generative model is employed as powerful prior for the underlying artwork, while crack structures are captured using a Mumford--Shah-type variational functional together with a crack prior. Joint optimization yields a pixel-level map of crack localizations in the painting.