Towards automated and surface adaptive Reflectance Transformation Imaging for cultural heritage artefacts

Photo above: LightBot set up – A robotic arm based RTI acquisition system

Towards automated and surface adaptive Reflectance Transformation Imaging for cultural heritage artefacts

Photo: Illumination directions and the robot motion visualised in ROS environment

Reflectance Transformation Imaging (RTI) is a well-established technique used for characterizing the surface geometry. It is particularly popular in the field of cultural heritage objects conservation/restoration, thanks to affordable setups and available software for visual examination of surfaces through relighting. Acquisition, modelling, and visualization are the three components of RTI. The acquisition process consists of capturing a set of images of an object with a fixed camera position while varying illumination directions. In general, light source(s) are manually handled, set in fixed positions (dome) or in best case positioned by an automated system. In all these cases, because there is no adaption to the observed objects, the data (acquired images) are of lower quality which impacts the modelling and visualization steps negatively. In the current state of the art, there are no methods for automatic planning of Light Positions (LPs). To deal with this problem, Early Stage Researcher Ramamoorthy Luxman developed progressive-iterative method to automatically estimate best LPs in an acquisition. The Next Best Light Position (NBLP) problem is to determine which LPs in an RTI acquisition process must be acquired to achieve the best adaption to the surface properties of an a-priori unknown object.

Current systems that are used to perform RTI acquisition consist mainly of free-form dome-based. The current state of these systems has many limitations like irreproducibility, inability to adapt to the size of the objects being acquired, speed and portability. We are developing a novel Robot-based system, as well as a Reflectance Transformation Imaging (RTI) associated framework, for the optimization of RTI data stitching in terms of acquired images and light positions. The proposed methodology allows the robust automation and reproducibility of series of acquisitions of large or complex scenes in a two-dimensional space, while optimizing pixel resolution.

The methods and system being developed has signi&cant practical usefulness to the cultural heritage objects conservators and restorers. You can read more in Luxman’s recent publication on the next best light position.