Review on texture descriptors for image classification

[abstract]

The goal of this chapter is to find empirically the best methods for describing a given texture using an ensemble to harness the discriminative power of different texture approaches. We begin our investigation by comparing the performance of a large number of different texture descriptors and their fusions. The best fusion approach is then tested across a diverse set of databases and compared with some of the best performing approaches proposed in the literature. Whenever possible the original code of each approach is used on the datasets for fair comparison. Both stand-alone and ensembles of texture descriptors are investigated. In addition, some tests based on deep learning features are reported. The support vector machine (SVM) is tested as a stand-alone classifier and as the base classifier in ensembles. Extensive experiments conducted on benchmark databases spanning several domains show that our proposed approach outperforms recent state-of-the-art approaches. The proposed tool is available at (https://www.dei.unipd.it/node/2357 + Pattern Recognition and Ensemble Classifiers).

Keywords Texture descriptors; Ensemble; local binary patterns; deep learning; Support vector machines

[full paper]