Gabor filter-based texture features to archaeological ceramic materials characterization

Abstract

This paper presents a self-learning system for automatic texture characterization and classification on ceramic pastes or fabrics and surfaces. The system uses Gabor filter as pre-processing methods with feature extraction possibilities. On these features it applies a linear discriminant analysis (LDA) and k-nearest neighbor classifiers (k-NN) with its best parameters. Experimental results of the recognition ceramic materials, deals on the field and in the laboratory, for different ceramic pastes and surfaces show a good accuracy and applicability of the process on this type of data. © 2012 Springer-Verlag.