On the role of logic in image retrieval

Image retrieval is a relatively new discipline, which has impetuously developed in the last few years. A distinguishing feature of this development is its simultaneous but separate arising in several sectors of computer science, such as Pattern Recognition, Vision, Information Retrieval, and Databases, to mention the most relevant. This fact reveals that there are many different aspects involved in image retrieval, each requiring a specific background and methodology to be successfully tackled, and also that there may be complementary approaches to the same problems, not only within the same discipline (such as different index structures for image data), but also cutting across different disciplines (such as similarity- \emph{versus} semantic-based image retrieval). This richness of paradigms, methods and systems may, on the long run, become a negative attitude, resulting in duplication of effort and loss of potentiality, ultimately prone to slow down progress. In connection with this problem, we discuss the role of logic in the image retrieval endeavor, arguing that logic has a twofold potentiality. First, it can be used to model \emph{specific forms} of image retrieval. Second, it is a most natural candidate to express a \emph{general} model of image retrieval, encompassing in a unique, formally specified and well-founded framework, the many forms of retrieval that have been studied so far in the areas mentioned above.