One can choose, on the other hand, to localize points on a three-dimensional rendering of the volumetric data. Surface rendering requires an explicit description of the surface that can be obtained by segmenting the data set, i.e., by identifying the voxels that belong to a given 3-D object. Here again, one can take advantage of the higher resolution of the video display to make the image smoother than the actual data.
Images acquired by magnetic resonance are subject to intensity inhomogeneity. Causes for this are identified by Simmons et al. [56] as non-uniform slice profile, bandwidth filtering of data, mistuning of the RF coils and crosstalk between contiguous slices. The negative effect of intensity inhomogeneity appears when intensity-based segmentation of the skin surface is performed prior to surface rendering and registration based on surface matching. A technique for correcting this problem is presented by Lim et al. [39].