Multiscale 3D edge representation of volume data by a DOG wavelet

This paper proposes a method to expand volume data into the 3D DOG (D#wence of Gaz~~s&ns) functions by using the frame theory of non-orthogonal wavclcts. The spherically symmetric fcaturc of the 3D DOG fnnction is suitable for the visualization methods based on the volume density projection. Since the DOG function approximates a V”G (Laplacian of Gaussian) fnnction, the rcprcscntation can bc consiclcrcd as a hierarchy of the 3D cclgcs on diffcrcnt scales. Thcrcforc WC can cnhancc the cdgc information at will by blending the projection images on cliffcrcnt scales. Since the wavclct cocfficicnts have significant value whcrc the volume dcnsity changes, WC may nsc this rcprcscntation method for the cnhanccmcnt of the biomedical fcaturcs and also can WC it as a data compression method by ncglccting the insignificant coctficicnts. WC will apply onr rcprcscntation method to medical CT volume data and show the cfficicncy in &scribing the spatial structnrc of the volume.