Automatic subarea selection for image registration
US5495540A · kind A · utility
Assignee
Inventors
Key dates
| Filing date | Oct 12, 1994 |
| Grant date | Feb 27, 1996 |
| Priority date | — |
| Expiry date | Oct 12, 2014 |
Classification
- Technology area (CPC G)Physics
- CPC primaryG06T7/337
- WIPO fieldComputer technology
- WIPO sectorElectrical engineering
Abstract
A method for automatically selecting subareas from reference image data such that registration accuracy is optimized between images. Optimal subarea selection reduces the on-line computation and reference data storage required for multi-subarea correlation or, alteratively, improves its effectiveness. Example results for a synthetic aperture radar (SAR) image are described. The results indicate that the automatic subarea selection method of the present invention reduces on-line computation by a factor of 2 to 3 (relative to random subarea selection) without degradation in accuracy. The present selection method minimizes the predicted total mean squared registration error (MSE). The total MSE is predicted in terms of the position and predicted measurement covariance (derived from local image statistics) of each candidate subarea. Combinatorial optimization procedures select a predetermined number of subareas to minimize total MSE. The present method jointly considers the quality and position for each local measurement, optimally handles any order polynomial model, and includes bounds on the uncertainty of registration parameters.
Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.