Global Semantic Classi cation of Scenes using Power Spectrum Templates

Scene recognition and content-based procedures are of great interest for image indexing applications processing very large databases. Knowing the category to which a scene belongs, a retrieval system may lter out images belonging to other categories. In this paper, we introduce a computational approach which classi es and organises real-world scenes along broad semantic axes. Fundamental to our approach is the computation of global spectral templates providing a continuous organisation of scenes between two categories. These templates encode the structure which is discriminant between two categories. We propose a hierarchical procedure of two stages, that organises images along three semantic axis. Firstly, all the scenes are classi ed according to an Arti cial to Natural axis. Then, natural scenes are organised along the Open to Closed axis whereas arti cial environments are classi ed according to the Expanded to Enclosed scenes axis.