Data fusion in 2D and 3D image processing: an overview

This paper presents an overview of the current state of the art in image fusion, with an emphasis on the emergence of new techniques, often issued from other domains like artificial intelligence and uncertainty modeling. We address the two following points: firstly the aim of data fusion and its specificity when image information has to be combined, with emphasis on the respective roles of numerical and symbolic information, vs. numerical and symbolic types of treatment, secondly the theoretical frameworks for modeling imprecision and uncertainty (probability, fuzzy sets, belief functions). The main steps of image fusion are illustrated in a simple example in 3D medical image fusion.