Rock Image Classification Using Non-Homogenous Textures and Spectral Imaging

Leena Lepistö
Tampere University
of  Technology
Institute of Signal Processing
P. O. Box 553, FIN-33101 Tampere




Keywords: Texture classification, Non-homogenous textures, Spectral Imaging, Rock Images


Texture analysis and classification are usual tasks in pattern recognition. Rock texture is a demanding classification task, because the texture is often non-homogenous. In this paper, we introduce a rock texture classification method, which is based on textural and spectral features of the rock. The spectral features are considered as some color parameters whereas the textural features are calculated from the co-occurrence matrix. In this classification method, non-homogenous texture images are divided into blocks. The feature values are calculated for each block separately. In this way, the feature values of the texture image can be presented as a feature histogram. The classification method is tested using two types of rock textures. The experimental results show that the proposed features are able to distinguish rock textures quite well.