Vassiliki Kokla

Ink Discrimination Based on Co-occurrence Analysis of Visible and Infrared Images

Inks found in Byzantine manuscripts are semi- transparent pigments and their examination and analysis provide an invaluable source of information on the authenticity and dating of manuscripts and the number of authors involved. However, inks are difficult to characterize because their intensity depends on the amount of liquid spread during scripting and the reflective properties of the support. Most existing methods for the analysis of ink materials are based on destructive testing techniques that require the physicochemical sampling of data. Such methods cannot be widely used because of the historical and cultural value of the manuscripts. In this work we show that manuscript inks can be represented through a mixture of Gaussian functions and can be characterised using co-occurrence matrices.

Revealing the Visually Unknown in Ancient Manuscripts with a Similarity Measure for IR-Imaged Inks

One of the tasks facing historians and conservationists is the authentication or dating of medieval manuscripts. To this end it is important to them to verify whether writings on the same or different manuscripts are concurrent. In this work we explore thistask by capturing images of manuscript pages in infrared (IR) and modelling and then comparing the ink appearance of segmented text. The modelling of the text appearance relies on the unsupervised multimodal clustering of ink descriptors and the derived probability density functions. The similarity measure is built around the distribution of cluster labels and their proportions. We demonstrate our method by using both model inks of known composition and authentic Byzantine manuscripts.

Unsupervised ink type recognition in ancient manuscripts

One of the tasks facing historians and conservationists is the authentication or dating of medieval manuscripts. To this end it is important to verify whether writings on the same or different manuscripts are concurrent. This work considers the problem of capturing images of manuscript pages in near-infrared (NIR) spectrum and compare the ink appearance of segmented text and their textural features. We present feature descriptors that capture the variability of the visual properties of the inks in NIR. Comparison of inks of unknown composition is achieved through unsupervised multi-dimensional clustering of the feature descriptors and similarity measures of derived probability density functions.

A Bayesian decision model for watercolour analysis

Bayesian Classification methods can be applied to images of watercolour paintings in order to characterize blue and green pigments used in these paintings. Pigments found in watercolour paintings are semi-transparent materials and their analysis provides important information on the date, the painter, the place of the production of watercolour paintings and generally on the authenticity of these works of art. However, watercolour pigments are difficult to characterize because their intensity depends on the amount of liquid spread during painting and the reflective properties of the underlying support. The method describedin this paper is non-destructive, non invasive, does not involve sampling and can be applied in situ. The methodology is based on the photometric properties of pigments and produce computational models which classify diverse types of pigments found in watercolour paintings. These pigments are photographed in the visible and infrared area of electromagnetic spectrum and models based on statistical characteristics of intensity values using a mixture of Gaussian functions are created. Finally the pigments are classified using a Bayesian classification algorithm to process the generate models.

Probabilistic image-based characterization of manuscript inks

Inks constitute the main element in Medieval manuscripts and their examination and analysis provides an invaluable source of information on the authenticity of the manuscripts, the number of authors involved and dating of the manuscripts. Most existing methods for the analysis of ink materials are based on destructive testing techniques that require the physicochemical sampling of data. Such methods cannot be widely used because of the historical and cultural value of manuscripts. In this paper we present a novel approach for discriminating and identifying inks based on the correlations of image variations under visible and infrared illumination. Such variations are studied using co-occurrence matrices and detect the behavior of the inks during the scripting process.

Near-Infrared Ink Differentiation in Medieval Manuscripts

One of the tasks facing historians and preservationists is the authentication or dating of medieval manuscripts. To this end it is important to verify whether writings on the same or different manuscripts are concurrent. We propose a novel approach for the automated image-based differentiation of inks used in medieval manuscripts. We consider the problem of capturing images of manuscript pages in near-infrared (NIR) spectrum and compare the ink appearance and textural features of segmented text. We present feature descriptors that capture the variability of the visual properties of the inks in NIR based on intensity distributions of histograms and co-occurrence matrices. Our approach is novel as it is entirely image based and does not include the spectrum analysis of the inks. The method is validated by using model ink images manufactured based on known recipes and ink segmented from medieval manuscripts dated from the 11th to the 16th century. Model inks are classified by using both supervised and unsupervised clustering. Comparison of inks of unknown composition is achieved through unsupervised multi-dimensional clustering of the feature descriptors and similarity measures of derived probability density functions.

An ink texture descriptor for NIR-imaged medieval documents

In this work we explore the task of authenticating and dating ancient manuscripts by capturing images of pages in nearinfrared (NIR) and modelling and then comparing the ink appearance of segmented text. We present a texture feature descriptor to characterize and recognize semi-transparent materials such as the inks found in manuscripts. These textural patterns are different in nature from perceptual entities such as textons, tokens, frequency or repeatability of textural elements. Our ink texture descriptor relates a set of ink features from various first and second-order statistics to semi-liquid and viscous image-based properties of inks. In particular, we propose eigen features from the joint gray-level probabilities and off-diagonal sums of co-occurrence matrices. We test the qualities of the features with a classifier trained with the ink descriptor to show how well it recognizes eight different inks of known composition. Presented with the very same task the human visual system would fail to spot the ink composition difference given the inks inter-class and intra-class distances are extremely short.