Nonlinear anisotropic diffusion filters are iterative, ``tunable'' filters introduced by Perona and Malik [39]. Gerig et al. used such filters to enhance MR images [17]. Sapiro and Tannenbaum used a similar technique to perform edge preserving smoothing of MR images [44]. In the extreme case, such smoothing might produce a profile of RF inhomogeneity in the images. Others have shown that diffusion filters can be used to enhance and detect object edges within images [1][39][34].
Since these filters smooth or enhance MR images and detect edges, they might also be used for RF correction and/or intracranial boundary detection in MR images. To this end, nonlinear anisotropic diffusion filters are studied here.