Automatically generating keywords for georeferenced images

The possibility of automatically annotating information objects such as images with metadata related to their geography is increasingly relevant, as larger volumes of such information are captured with explicit georeferences (e.g. Flickr images). Within the Tripod project research has been carried out as to the extent to which we can automatically derive keywords and captions for images which are captured by cameras capable of automatically storing coordinates and azimuth information, that is to say not only the position from which a picture was taken, but also the direction in which the camera was pointing. Improved tagging of images can improve search, and particularly in commercial contexts, reduce annotation costs which are an important part of the work of image libraries (Iwasaki et al., 2008).