IG03 - Research Group for Computer Science and Human-Computer Interaction

Research in this area focuses on developing advanced algorithms for the automatic restoration of damaged historical photographs. Generative image synthesis methods, including GANs and diffusion models, are used to simulate complex physical damage and create robust datasets for training AI systems. This approach enables the development of highly efficient methods for reconstructing missing image parts (inpainting) and for global image reconstruction and restoration that are robust to various forms of degradation. The main objective is to create specialised datasets and models that ensure accurate digital restoration while maximising the preservation of the historical authenticity of visual heritage.