Content-Based Retrieval for A Database of Function Approximated Kamon Images

In this paper, a content-based retrieval method for a database of Japanese kamon images is introduced. Kamons are traditional family emblems that have both commercial and historical research values. Our method is based on comparing the shape of geometric patterns in a pair of images by relaxation matching in which the objects and the labels are represented as function approximated contour segments. Images of closest similarity according to a scalar distance function are returned as output. Our goal is to include this method as an added value function in an existing software package based on kamons where the only available retrieval method is by keywords or pre-defined categories. This work is part of a government-funded research project to transfer technology from university to industry in order to increase its competitiveness. Experimental results on a database of 2,000 kamon images are included for evaluation.