Computers analyzing old masterpieces

Abstract

This lecture deals with the usage of digital image processing methods in cultural heritage applications. Several main goals and issues will be discussed and examples of solutions will be described. Following groups of methods will be addressed: data acquisition, where recent technology progress has enabled hyperspectral representation of artworks and new methods for the data analysis and processing are needed; data preprocessing, which increases the quality of the acquired dataset, using deblurring, denosing or even removal of more complex artefacts. Then, the mentioned higher level methods are change detection, trying to use multimodal dataset to uncover hidden changes, an authorship verification addressing possible forgeries using texture and high frequency based features, and geometry analysis, where the geometry of a painting scene is recreated and used for hypothesis disproval. The last covered topic will deal with the analysis of unseen data, having on mind microscopic data used during material analysis and data fusion from different modalities, where information acquired in non-visible spectra can bring information about hidden layers of a painting and thus recover original painter’s intentions.