Spherical diffusion for 3D surface smoothing

Area of Science:

• Computer Vision
• Computer Graphics
• Computational Geometry

Background:

• Surface smoothing is crucial in various scientific and engineering fields.
• Traditional methods like iterative smoothing can be computationally expensive and suffer from shrinkage artifacts.
• Representing surfaces as scalar functions on a sphere is a common approach in fields like medical imaging and geodesy.

Purpose of the Study:

• To present a novel diffusion-based approach for surface smoothing on the sphere.
• To develop a computationally efficient method that avoids iterative processes and shrinkage.
• To demonstrate the preservation of important shape features and introduce a modification for improved smoothing of challenging objects.

Main Methods:

• Surfaces are represented as scalar functions defined on the sphere.
• A diffusion process, equivalent to Gaussian smoothing on the sphere, is employed.
• A nonlinear modification of the diffusion process is introduced for enhanced smoothing of specific object types.

Main Results:

• The diffusion-based approach is computationally efficient, avoiding iterative smoothing.
• The method does not exhibit the shrinkage problem common in other smoothing techniques.
• The evolution of parabolic curves, representing important shape features, under diffusion is successfully demonstrated.
• The nonlinear modification improves smoothing for elongated and poorly centered objects.

Conclusions:

• The presented diffusion-based method offers an efficient and effective solution for surface smoothing on the sphere.
• This approach preserves critical shape characteristics and overcomes limitations of traditional methods.
• The nonlinear modification enhances its applicability to complex surface geometries.