Self-Organizing Map with Refractoriness and Its Application to Image Retrieval

Abstract

In this research, we propose a Self-Organizing Map with Refractoriness (SOMR) and apply it to a similarity-based image retrieval. The proposed SOMR is based on the Self-Organizing Map (SOM) and the refractoriness which is observed in the real neuron is introduced. In the proposed SOMR, the plural neurons in the Map Layer corresponding to the input can fire sequentially because of the refractoriness. The image retrieval system using the SOMR makes use of this property in order to retrieve plural similar images. In this image retrieval system, as the image feature, not only color information but also impression words and key words are employed. We carried out a series of computer experiments and confirmed that the effectiveness of the proposed SOMR and the image retrieval system using the SOMR.