A novel 3D anisotropic diffusion filter
I. Pop, Romulus Terebeş, Jean‐Pierre da Costa, Christian Germain, Monica Borda, Cosmin Ludusan, Olivier Lavialle · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2010
We propose a 3D anisotropic diffusion filter for denoising and enhancing oriented data. The approach is a generalization of a previous contribution and it is based on a novel, more precise orientation estimation step. The orientation in the 3D space is computed using an asymmetric Isotropic Recursive Oriented Network (IRON) operator that can handle in a natural way junctions and corners. In the experimental section we employ a set of 3D synthetic blocks to illustrate the efficiency of the new method through quantitative and visual comparisons with other 3D-extended classical models or recently proposed 3D Partial Differential Equations (PDE).