Digital Restoration of Ancient Mosaics
I've always longed to witness the restoration of ancient mosaics, especially the iconic The Gypsy Girl mosaic from the Zeugma Museum. Physical restorations have gained notoriety over the last few years due to some unfortunate incidents that have occurred. As a result, performing digital restorations emerge as a more practical approach, providing the advantage of avoiding potential catastrophic outcomes. Although there already exists a body of research that investigates the formations and colors of tiles, mostly for indexing purposes, digital restoration methods have not fully matured over the years. In 2018, I gained experience in developing a manual restoration tool that utilized techniques such as adaptive histogram equalization (CLAHE), water-shedding, tile segmentation, k-NN clustering, and interpolation. However, this process proved to be exceedingly laborious, time-consuming, and prone to errors. Thus, I envisioned that generative AI methods would play a crucial role in streamlining such digital restoration processes.
With the inclusion of theGenerative Fill feature, performing digital restoration of ancient mosaics has become significantly more convenient. I tried out this method on several images of damaged mosaics, and I must say that I'm genuinely impressed with the results it produces. The improvement achieved is beyond what I initially anticipated. It has proven to be an incredibly valuable tool for the restoration process, effectively revitalizing the broken mosaics with remarkable accuracy and attention to detail. As a next step, I intend to delve deeper into developing custom generative AI methods exclusively for digital restoration purposes. My goal is to further enhance and refine the restoration outcomes achieved through this tool.
The Gypsy Girl
Deesis (Haghia Sophia)
Metiochus and Parthenope
Oceanus and Tethys
Daedalus
Clio and Euterpe
Dionysos and Ariadne
Dionysos Triumph
Other excavations
Sources: