Similarity-Based Image Retrieval Considering Size and Color of Artifacts by Self-Organizing Map with Refractoriness

Abstract

In this paper, we propose a similarity-based image retrieval considering size and color of artifacts by self-organizing map with refractoriness. In the proposed system, five search requests (retrieval of (1) images which are similar to key image, (2) images which are similar to a part of key image, (3) images which have similar feature of same positions, (4) images which include similar artifacts in different position and (5) images which have similar features to common features in plural key images) are considered. In the proposed system, not only color of natural objects, color of artifacts, spectrum and keyword but also size and color of artifacts are used as the image feature. As a result, we can obtain more desirable similar images to the key image(s). We carried out a series of computer experiments and confirmed that effectiveness of the proposed system.