I am currently testing FREAK descriptors in Matlab. To detect feature points, I used detectBRISKFeatures(), as suggested in the FREAK paper. To build FREAK descriptors, I used [features, validPoints]=extractFeatures(I, points, 'Method', 'FREAK'). Then, the descriptors of two images are matched.
However, when testing the Affine Covariant Regions datasets (from http://www.robots.ox.ac.uk/~vgg/data/data-aff.html), I found that the matching accuracy is much worse than SIFT. But, quite a few existing literature have shown that FREAK outperforms SIFT. Anyone knows what happens here? Or is there any problem in my matlab implementation of FREAK?
Thanks,
Guohua
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Control Theory
- Control Systems
More Guohua Lv's questions See All
Similar questions and discussions