Information-theoretic analysis of steganalysis in real images
MM&Sec '06 Proceedings of the 8th workshop on Multimedia and security
Limits of Learning-Based Superresolution Algorithms
International Journal of Computer Vision
| Hi-index | 0.00 |
This paper introduces a new generalization of scale-space and pyramids, which combines statistical modeling 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. Examples show how MGMM can be applied to problems such as segmentation and motion analysis.