IMAGE AND FACE ANALYSIS FOR PERSONAL PHOTO ORGANIZATION

In recent years, digital cameras are becoming very commonplace and users need tools to manage large personal photo collections. In a typical scenario, a user acquires a certain number of pictures and then transfers this new photo sequence to his PC. Thus, before being added to the whole personal photo collection, it would be desirable that this new photo sequence is processed and organized. For example, users may be interested in using (i.e., browsing, saving, printing and so on) a subset of stored data according to some particular picture properties. For these reasons, automatic techniques for content-based description of personal photos are needed. Tools enabling an incremental organization of the photo album should take advantage from the particular properties of digital library. Indeed, personal photo collections show peculiar characteristics as compared to generic image collections, namely, a relatively small number of different individuals can be detected across the whole collection and, generally, it is possible to group the photos based on specific attributes. In such a scenario, the user is mainly interested in who is in the picture and where and when the picture was shot. Considering Who, where and when as the fundamental aspects of photo information, in this thesis it will be given a detailed description of novel approaches for content-based image retrieval. Novel image analysis techniques will be presented, focusing in particular on face information. Then, two novel frameworks for personal photo organization will be shown.