SISREC: a system for image retrieval

An important problem in image management is image retrieval. When images in a database are not well organized, their efficient retrieval is a problem. Most of the existing systems use image-features to form the queries to the system. For the final user these features are difficult to use or do not have any meaning. The results output by the system most of the times can be unpredictable. Trying to reduce this problem, other systems use images as examples to get the results. Some others use icons. A few of them use the content to form the queries. In this note we describe a system called SISREC that falls in this category. It is able to retrieve images from a previously organized database in terms of their objects. SISREC is composed of three modules. The first module performs object training and object recognition. The second module performs image analysis and image organization. The third module allows for image retrieval. Through several experiments we show the efficiency of the system.