We propose in this chapter a content-based information retrieval framework to select documents in a database, consisting of several images with semantic information. Information in these documents is not only heterogeneous, but also often incomplete.