Image indexing by spatial relationships between salient objects

In this paper, we presented our technique to extract and to use the spatial relationships between two or more salient objects. Using an object oriented hypergraph data structure, the spatial relationships are determined and stored in an object-oriented database. This work aims to unified the phases of processing, indexing and retrieval of images. The proposed model can be applicable to other types of data (video) and to semantic relations hidden in an image. The structure of the database used for image storage allow the construction of the indexes classes hierarchy in order to improve the results of image retrieval. Our method requires more experiments for datasets which come from different areas and for images which contain more salient objects.