During the past 5 years, there has been a considerable effort of research in automating the analysis and fusion of multidimensional images, yielding important theoretical and practical new results. These results could now be useful to brain research. Along these lines, and focusing on 3D images of the brain obtained with CT, MRI, SPECT, and PET techniques, I will present a number of methods useful to produce quantitative measurements necessary for an objective analysis of 3D images of the brain. Such methods include segmentation, shape analysis, rigid and elastic registration, fusion of multimodal images, analysis of temporal sequences (4D data), modeling, and matching of digital atlases. A large number of references on these topics can be found in Ayache (1995a,b), and specific examples from our research group in Malandain et al. (1994), Subsol et al. (1995), and Thirion (1995). The following text describes the research tracks to be followed in order to optimize the future exploitation of 3D images of the brain in neuroscience.