Stochastic Processes in Vision: From Langevin to Beltrami

Diffusion processes which are widely used in low level vision are presented as a result of an underlying stochastic process. The short-time non-linear diffusion is interpreted as a Fokker-Planck equation which governs the evolution in time of a probability distribution for a Brownian motion on a Riemannian surface. The non linearity of the diffusion has a direct relation to the geometry of the surface. A short time kernel to the diffusion as well as generalizations are found.

• Type: Conference
• Year: 2001
• Where: ICCV
• Authors: Nir A. Sochen
• Keywords: Computer Vision, ICCV 2001, Low Level Vision, Non Linearity, Riemannian Surface, Short Time Kernel, Short-time Non-linear Diffusion