Method and apparatus for applying Gaussian Mixture Models to local image patches using an adaptive color lookup table
US8290253B1 · kind B1 · utility
Assignee
Inventors
Key dates
| Filing date | Oct 30, 2009 |
| Grant date | Oct 16, 2012 |
| Priority date | — |
| Expiry date | Dec 17, 2030 |
Classification
- Technology area (CPC G)Physics
- CPC primaryG06V10/763
- WIPO fieldComputer technology
- WIPO sectorElectrical engineering
Abstract
A method, system, and computer-readable storage medium for applying Gaussian Mixture Models (GMMs) to local image patches using an adaptive color lookup table. Per-channel color quantization may be performed to find representative colors for a local image patch. Each combination of the representative values corresponds to a representative color. The probabilities of the representative colors may be computed using a local GMM color model and stored to corresponding entries in an adaptive color lookup table. For every pixel in an image patch, the closest representative color may be found, and the corresponding probability may be retrieved from the lookup table and used for the pixel. The method may, for example, be applied to each local window in a method for automatically determining segmentation in a digital video image sequence to calculate the foreground probabilities for the pixels in a propagated classifier via a GMM.
Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.