Characterization of Isothetic Polygons for Image Indexing and Retrieval

A novel algorithm for two-level indexing of a binary image in a database is proposed. The scheme is based on certain elementary geometric features of the tightly fitting isothetic polygon circumscribing the digital object (and/or those inscribing the holes in the interior of the object). These features, being simple and precise, are used to form a higher level description of the underlying objects present in an image. The edges of an isothetic polygon are classified into three types based on the types of vertices defining an edge. The number of each type of edges is used for first-level of indexing, and the total edge length is used in the second level. The scheme has been tested on two benchmark logo databases, and the retrieval results demonstrate the elegance, robustness, efficiency, and speed of the proposed method