Classification in Big Image Datasets Using Layered-SOM

Adequately classifying big image datasets containing images of arbitrary domains is getting more and more important nowadays. However the above mentioned problem has yet to be solved generally. The most suitable descriptors recognizing possible underlying structures and similar characteristics within large image datasets have to be selected and combined in order to carry out multi-feature analysis and thus image classification. This paper presents an enhancement of the original SOM via developing an unsupervised learning method using multiple layers. This method is appropriate of analyzing and classifying big image datasets with combining multiple image descriptors nonlinearly. It increases the precision of image clustering as well as reducing the time required for computation.