Texture Image Classification using Unsupervised Learning

Nedyalko Petrov and Ivan Jordanov


Pattern recognition and classification, unsupervised learning, self-organizing maps, feature extraction


We investigate further our intelligent machine vision system for pattern recognition and texture image classification. A database of about 335 texture images of industrial cork tiles is used for this research. The images need to be classified into several classes based on their texture features similarities. In this work, we assume that there is no a priori human vision expert knowledge about the classes. After pre-processing of the data, feature extraction and conducting statistical analysis by applying principal component analysis (PCA) and linear discriminant analysis (LDA), we investigate unsupervised neural network learning. Self-organizing map (SOM) neural networks are trained, tested and validated and the obtained results are discussed and critically compared with research works investigating similar approaches.

Important Links:

Go Back