A Similarity Measuring Method for Images Based on the Feature Extraction Algorithm using Reference Vectors

We propose a similarity measuring method for images based on our feature extraction algorithm. The method represents features of an image as a feature vector called reference vector which is a relative measure extracted indirectly from images while many of existing methods use an absolute similarity measure extracted directly from images. A reference vector is calculated from correlation matrices of an image and reference images. Considering reference images as axes of a coordination system, our method enabled users to extract their intended features by selecting appropriate images as reference images. This significant characteristic of the method is effective to measure similarity based on users' preference and to differentiate an image from others. In this paper, we illustrate our method in detail and demonstrate its effectiveness through experiments.