Allure of Craquelure: A Variational-Generative Approach to Crack Detection in Paintings
Researchers have developed a novel hybrid approach for detecting cracks in paintings, crucial for art conservation and documentation. This method models crack detection as an inverse problem, separating the observed image into a crack-free painting and a crack component. A deep generative model acts as a prior for the artwork, while a Mumford--Shah-type variational functional and a crack prior capture the crack structures. Joint optimization then produces a precise pixel-level map of crack localizations. AI