Historical Manuscript Dating using Textural Measures
Anmol Hamid, Maryam Bibi, Imran Siddiqi, Momina Moetesum · 2018
Historical manuscript dating is a challenging task that requires extensive domain knowledge regarding varying scripts, writing styles and instruments involved. Increased digitization of historical documents for preservation and analysis purposes have encouraged the relevant community to design techniques for computerized dating of historical manuscripts. This paper presents a performance based comparative analysis of popular textural features for document dating purposes. Outcomes of a comprehensive series of experiments demonstrate that a combined textural features-based approach is more effective than employing individual texture features. A 345-dimensional feature vector is extracted by using a combination of individual features from Gabor filters, Uniform Local Binary Patterns and Histogram of Local Binary Patterns, which is then fed to a Linear Discriminant Analysis (LDA) classifier. Performance evaluation of our proposed scheme on the Medieval Paleographical Scale (MPS) dataset gives a reduced Mean Absolute Error of 20.13.