Automatizing chromatic quality assessment for cultural heritage image digitization

In the context of digitization of photographs and other documents with graphical value, cultural heritage organizations need to give a guarantee that the stored digital image is a faithful representation of the physical image both at the physical level and the perceptual level. On the physical level, image quality can be measured objectively in a simple way by applying certain physical attributes to the image, as well as by measuring how distorting images affects the performance of the attributes. However, on the perceptual level, image quality should correspond to the perception that a human expert would experience when observing the physical image under certain determined and controlled conditions. In this paper we address the problem of image quality assessment (IQA) in the context of cultural heritage digitization by applying machine learning (ML). In particular, we explore the possibility of creating a decision tree that mimics the response of an expert on cultural heritage when observing cultural heritage images.