Chandrajit Bajaj
 Introduction | Visualization Tool | Filtering | Gradient Vector Diffusion | Segmentation | Skeletonization | References

4. 3D Gradient Vector Diffusion

We propose a new method for gradient vector diffusion, based on anisotropic PDE-based diffusion:

where g(alpha) is a decreasing function and alpha is the angle between the central vector and the surrounding vectors. For faster implementation, the calculation of the angle between two vectors is usually approximated by the inner-product of two vectors divided by their magnitudes.