An Adaptive Gauss Filtering Method

An adaptive filtering method for volume data is presented in this paper. In this filtering method, the input data set is re-sampled to create a hierarchy of multiple-level data sets. A data classification task is performed at each level of the data pyramid to decide the local structure types. Data voxels are classified as
linear
,
planar
, or
blob
structures, based on the gradients and the eigenvalues of
Hessian

matrices
. The classification results are used to adjust the shapes and orientations of filters such that noises are suppressed while key features are preserved.