Medieval Coins of Three Different Types and of Various States of Preservation

Abstract

We have developed a device for digitizing coins using photometric stereo, which serves two purposes. For inventory it allows identifying a coin, which has been digitized before, and avoids mixing up similar coins. This is important because the classic marking directly on the object is not possible without obscuring the design. Secondly, one can view a digitized coin on screen and interactively change the light direction similar to Reflectance Transformation Imaging (RTI). This enables researchers to better recognize details, especially in the case of often corroded coin finds, and also enables location independent investigations and exchanges. The digitization result consists of color (albedo) and normal information for each pixel, which allows to analyze topographic properties apart from color. We think that this type of data can enable the development of new algorithmic analysis methods. The classification of coins, especially medieval coins, requires specialist knowledge and a great deal of experience. Digital support can help archaeologists without numismatic knowledge to classify coins correctly by providing initial clues and showing, which coins in a comparative data base show similarities with a newly found coin. For the development of such digital tools, we provide a selection of coin data as an open dataset. For the dataset we have selected samples of medieval coins from three different types, which are described in Mehl 499, 595 and Bahrfeldt 19. The dataset contains 2D and 3D data only for their obverses. A possible research direction could be to measure similarity between these samples, such that samples of the same type are more similar than samples of different type. Many samples show only a part of a complete coin. This increases the challenge e.g. for shape correspondence. Multi-scale integral invariant (MSII) features included with the 3D data may help to focus on minting features.

Keywords

• numismatics
• medieval coins
• multi-light single-camera digitization
• surface normal
• 3d data
• multi-scale integral invariant features
• similarity