Multiresolution Gaussian mixture models for visual motion estimation

This paper introduces a new generalisation of scale-space and pyramids, which combines statistical modelling with a spatial representation. The representation uses the familiar concept of multiple resolutions, but applied to a Gaussian mixture representation of the image - hence the title MGMM. It is shown that MGMM can approximate any probability density and can adapt to smooth motions. After a brief presentation of the theory, it is shown how MGMM can be applied to the estimation of visual motion.

• Type: Conference
• Year: 2001
• Where: ICIP
• Authors: Roland Wilson, Andrew Calway
• Keywords: Gaussian Mixture Representation, ICIP 2001, Image Processing, Probability Density, Spatial Representation