A volume data classification method based on 3D spherical filters

This paper describes a method to classify 3D volume data by a technique based on fractal dimension analysis. To estimate fractal dimension of the 3D solid textures, the Hurst analysis technique is applied to 3D volume data for classifying texture patterns. In the analysis, popular two dimensional circular Hurst operators are extended to three dimensional spherical operators. The extension of the Hurst operator makes it possible to extract pattern features directly from 3D volume data, whereas typical circular Hurst operators extract pattern features only from 2D image data. An experimental database of 3D volume data is synthesized by Perlins' noise functions. The database is used for testing three dimensional spherical Hurst operators. Preliminary experimental results show the efficiency of the technique for classifications and segmentations of 3D volume data.