Abstract
In order to solve the problem of matching two images with different resolutions, a one-to-many image matching strategy is proposed, employing the Gaussian image pyramid. Modified scale invariant feature transform (SIFT) descriptor is used to extract the image feature of the local region around each corner point, and used to match the corner. A simple and robust method based on sample consensus is presented to eliminate spurious matches. After all the spurious matches are eliminated, the best one-to-one image matching result is chosen from all the matching results as the final exact result. Extensive experiments showed that the proposed method is able to match images with different resolutions effectively, and that it is robust and easy to compute.
| Original language | English |
|---|---|
| Pages (from-to) | 598-601 |
| Number of pages | 4 |
| Journal | Beijing Ligong Daxue Xuebao/Transaction of Beijing Institute of Technology |
| Volume | 28 |
| Issue number | 7 |
| Publication status | Published - Jul 2008 |
Keywords
- Feature descriptor
- Gaussian image pyramid
- Image matching