Fusion in image processing: specific features and digital approaches

Information fusion groups together the techniques used to combine various types of information on the same problem. In image processing, information fusion is concerned with how best to combine images of different origins to gain a better understanding of the object under observation. Fusion has become an important aspect of information processing in several very different fields, in which the information to be fused, the objectives, the methods, and therefore the terminology, can vary greatly, even if the analogies are also numerous. The scope of information fusion is keeping pace with that of technology and information processing in general.

The aim of this article is to clarify the context and concepts of fusion in the field of signal and image processing (SIP), to provide definitions and to outline the main digital approaches. Rule-based, syntactic, logical and neural approaches are not presented here.

Paragraph 1 presents a general definition, the characteristics of the data to be taken into account in a fusion system, and the main steps involved. Paragraph 2 is devoted more specifically to the specifics of fusion in image processing, highlighting what distinguishes it from fusion in other fields. The main numerical approaches are then outlined, in paragraphs 3 for probabilistic and Bayesian approaches, 4 for belief function theory, and 5 for fuzzy and possibilistic methods. Finally, in paragraph 6, we discuss the processing of spatial information in image fusion.