Ink recognition based on statistical classification methods

Statistical classification methods can be applied to images of historical manuscripts in order to characterize the various kinds of inks used. As these methods do not require destructive sampling they can be applied to the study of old and fragile manuscripts. Analysis of manuscript inks based on statistical analysis can be applied in situ, to provide important information for the authenticity, dating and origin of manuscripts. This paper describes a methodology and related algorithms used to interpret the photometric properties of inks and produce computational models which classify diverse types of inks found in Byzantine-era manuscripts. Various optical properties of these inks are extracted by the analysis of digital images taken in the visible and infrared regions of the electromagnetic spectrum. The inks are modelled based on their grey-level and colour information using a mixture of Gaussian functions and classified using Bayes' decision rule