A Novel Real-Time Feature Matching Scheme
Affine Scale Invariant Feature Transform (ASIFT) can obtain fully affine invariance, however, its time cost reaches about twice that in Scale Invariant Feature Transform (SIFT). We propose an improved ASIFT algorithm based on feature points in scale space for real-time application. In order to detec...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IFSA Publishing, S.L.
2014-02-01
|
| Series: | Sensors & Transducers |
| Subjects: | |
| Online Access: | http://www.sensorsportal.com/HTML/DIGEST/february_2014/Vol_165/P_1890.pdf |
| Summary: | Affine Scale Invariant Feature Transform (ASIFT) can obtain fully affine invariance, however, its time cost reaches about twice that in Scale Invariant Feature Transform (SIFT). We propose an improved ASIFT algorithm based on feature points in scale space for real-time application. In order to detect the affine invariant feature point, we establish a second-order difference of Gaussian (DOG) pyramid and replace the extreme detection in the DOG pyramid by zero detection in the proposed second-order DOG pyramid, which decreases the complexity of the scheme. Experimental results show that the proposed method has a big progress in the real-time performance compared to the traditional one, while preserving the fully affine invariance and precision.
|
|---|---|
| ISSN: | 2306-8515 1726-5479 |