Self-organizing Map (SOM) Based Data Navigation for Identifying Shape Similarities of Graphic Logos

摘要

In this paper, we propose a data navigation approach for identifying the shape similarity of graphic logo images using enhanced SOM based visualization methods. Existing SOM based visualization methods have the limitation of not being able to show detailed local distance information and global similarity of data at the same time. Therefore, we propose two visualization approaches to overcome this limitation for better image data navigation. In our experiments, we used MPEG-7 shape image dataset and classic IRIS dataset to demonstrate our approaches are superior to previous approaches in providing sufficient local and global information for data visual navigation.