hello all,
I want to know what is the best corner detector ?
I read that harris is the best but i found some where that it exist an other method called CSS wich is better.
is it true and is there an other detector better ?
otherwise, i need an article as a reference about this.
Thank you
best regards
Hi,
I assume you are talking about computer vision.
Most of the recent articles require a subscription. I think this one will be useful for your purposes, though. You can find some articles if you look by CSS full name too (Curvature Space Scale)
Best,
Fiorella
http://ceng.anadolu.edu.tr/CV/CornerDetection/CornerDetection.pdf
Harris is the traditional and most reliable corner detector. However, other popular corner detector include SUSAN which trades off accuracy for speed, and FAST which greatly improves on the accuracy of SUSAN while maintaining high speed. most of the recent literature do not rank CSS with other corner detectors, but i think the decision to include CSS shoukd depend on your research direction.
thank you sir Abass Olaode
can you see this paper that compar
harris, nobel susan and CSS
and obtain that css is the best
please just see it and tell me if i can take this on cosidiration
sir Ahmad Hassanat I didn't get your question ?
The pleasure is mine Hadjer. I noticed FAST was not included in the study, but overall I think it is safe to pick any of the algorithms most applicable to your study.
Based on my experiments (check the publications) If no scale/rotation invariance is necessary, FAST and Harris and Shi/Tomasi are really good choices, with BRIEF for matching. FAST+BRIEF will be really fast, the others are a bit slower, but more accurate. For scale/rotation invariance, go with ORB or SIFT. SIFT will be a bit more accurate, but much slower.
The determinant of the Hessian is a very good interest point detector, with better performance than the Laplacian operator or difference of Gaussians. For theoretical and experimental results about this operator and other closely related differential interest point detectors, please see
Lindeberg (2013) ``Image matching using generalized scale-space interest points", Journal of Mathematical Imaging and Vision, Special issue with selected papers from SSVM 2013, in press,
http://link.springer.com/article/10.1007%2Fs10851-014-0541-0
http://www.csc.kth.se/~tony/abstracts/Lin13-SSVM.html
Lindeberg (2013) ``Scale selection properties of generalized scale-space interest point detectors", Journal of Mathematical Imaging and Vision, volume 46, number 2, pages 177-210.
ftp://ftp.nada.kth.se/CB/papers/Lin13-JMIV.pdf
I recommend you the corner detection of Harris (and Stephens). It is simple and fast...
http://www.bmva.org/bmvc/1988/avc-88-023.pdf
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Signal Processing
- Image Processing