Content based image retrieval using query by approximate shape

In this paper we present a new method for content-based image retrieval. The method is based on querying database by approximate shape representing given object. In this way all images containing the object may be found. Shapes are specified as a set of geometric primitives and attributes. Relations between primitives are represented by a graph. Our graph matching algorithm is used for computing the level of similarity between shapes. The method may be also used for searching transformed as well as partially covered objects. Experimental results showed the efficiency of our approach.