In the last years a great deal of content-based techniques for image indexing and retrieval has been developed. In spite of such important efforts, automatic seeking of images in a very large database is still a challenging task. In this paper we describe a strategy to cope with efficient indexing for query by example processing, which brings together Animate Vision and data discovery techniques. In the proposed method, images are represented in a novel manner by means of their Information Path so that hidden semantic associations among them, in terms of image categories, are discovered. Such associations eventually allow an automatic pre-classification, which makes query processing more efficient and effective.