Automatic Image Indexation to Support Content-Based Retrieval

This article presents a new approach for the automatic interpretation of the content of graphical and pictorial images and the creation of access structures based on this analysis. The automatic indexation of images is necessary to support the access to a database of images, giving conditions on the semantic image content. The key aspect of the image analysis process is on the composition of objects, in terms of other objects, taking into account different interpretations and different levels of recognition. The image analysis process can be applied to images belonging to specific domains, described in advance to the system. Image analysis is performed in two steps: in the first step, simple objects are recognized starting from basic graphical/pictorial elements; in the second step, complex objects are recognized (applying domain rules) as compositions of simpler objects and different interpretations are generated. The image query language allows the user to express conditions on the objects to be found in the images, on their relative positions on their interpretation and degree of recognition. A comprehensive example, based on 2D graphical images, is discussed, and the implemented prototype is described.