Efficient Osmosis Filtering of Thermal-Quasi Reflectography Images for Cultural Heritage
For cultural heritage restorers, this provides a computationally efficient method to correct light inhomogeneities in TQR images, enabling better analysis of subsurface structures.
The authors apply an image osmosis model to correct light inhomogeneities in Thermal-Quasi Reflectography images of cultural heritage, using efficient operator splitting to reduce computational cost. They demonstrate the method on a mural painting by Leonardo Da Vinci, producing a light-balanced image that can be registered to a visible orthophoto for restoration.
In Cultural Heritage, non-invasive infrared imaging techniques are used to analyse portions of deep structures behind wall paintings. When mosaicked, these images usually suffer from light inhomogeneities due to the experimental setup, which may prevent restorers from distinguishing the physical properties of the object under restoration. A light-balanced image is therefore essential for inter-frame comparisons, while preserving intra-frames details. In this paper we apply the image osmosis model proposed in (Weickert, 2013) to solve the light balance problem in Thermal-Quasi Reflectography (TQR) imaging. Due to the large amount of image data, the computation of the numerical solution of the model may be prohibitively costly. To overcome this issue, we make use of efficient operator splitting techniques. We test the proposed numerical schemes on the TQR measurement dataset of the mural painting "Monocromo" by Leonardo Da Vinci at Castello Sforzesco (Milan, Italy). The light corrected result is registered to a visible orthophoto, which makes it re-usable for further restorations.