Visual Clustering Of Trademarks Using A Component-Based Matching Framework

This paper describes the evaluation of a new component-based approach to querying and retrieving for visualization and clustering from a large collection of digitised trademark images using the self-organizing map (SOM) neural network. The effectiveness of the growing hierarchical self-organizing map (GHSOM) has been compared with that of the conventional SOM, using a radial based precision-recall measure for different neighbourhood distances from the query. Good retrieval effectiveness was achieved when the GHSOM was allowed to grow multiple SOMs at different levels, with markedly reduced training times.