Comparison of Sift,surf,orb,fast feature descriptor operators in OpenCV

Comparison of Sift,surf,orb,fast feature descriptor operators in OpenCV

Reference:

Http://wenku.baidu.com/link?url=1aDYAJBCrrK-uk2w3sSNai7h52x_ Eweru9p9ghzd49wj1beob7vluqdbdrkeehao2q3b7rattxdruq-m9cr-w2yqaterdliu1t3whyoyqfi

http://www.cvchina.info/2011/07/04/whats-orb/

Http://www.bubuko.com/infodetail-909956.html

The main features of the detection methods are the following, in the general image processing library (such as OpenCV, Vlfeat, BOOFCV, etc.) will be implemented.

• FAST, machine learning for high-speed Corner Detection, 2006
• Sift,distinctive image Features from Scale-invariant keypoints,2004, invariant to Image translation, Scaling, and rotation, partially invariant to illumination changes and robust to local geometric distortion
• surf,speeded up robust features,2006, inspired by sift, faster than sift, robust
• Orb,orb:an efficient alternative to SIFT or surf,2011, based on fast, two orders of magnitude faster than SIFT, which can be used as an alternative to SIFT
• Brisk,brisk:binary Robust Invariant Scalable keypoints
• Star,censure:center surround Extremas for realtime feature detection and matching, not a high number of citations
• Mser,robust Wide Baseline Stereo from maximally Stable extremal regions,2002, spot detection
• Gftt,good Features to Track,1994,determines strong corners in an image
• Harris,harris and M. Stephens (1988). "A combined corner and edge detector" is also a corner detection method .

The main concern here is the contrast of Sift,surf,fast,orb.

Image NO	SIFT	SURF	ORB	FAST
0	2414	4126	500	11978
1	4295	8129	500	16763
2	3404	4784	500	16191
3	1639	2802	500	7166
4	1510	1484	497	29562
5	10572	8309	500	720
6	191	187	295	16125
7	3352	4706	500	567
8	165	403	374	26701
9	4899	7523	500	12780
10	1979	4212	500	10676
11	3599	3294	500	36V
12	163	168	287	7923
13	1884	2413	500	11681
14	2509	5055	500	18097
15	9177	4773	500	7224
16	3332	3217	500	20502
17	5446	6611	500	16553
18	4592	6033	500	706
19	266	509	459	9613
20	2087	2786	500	7459
21st	2582	3651	500	12147
22	2509	4237	500	14890
23	1236	4545	500	6473
24	1311	2606	500	4293
25	237	387	500	657
26	968	1418	488	6609
Time Cost	21.52	17.4	0.97	0.25

You can see that fast extracts a large number of feature points, in the calculation time, faster than sift surf two orders of magnitude, the orb on fast basis, the quality of the feature point is higher!

Below through two pictures to see these algorithms match the effect, 1639-1311-697 represents the picture 1, 2 respectively extracted 1639,1311 keypoints, which matched 697.

Image pair	SIFT	SURF	ORB	FAST (SURF)
Eiffel-1.jpg,eiffel-13.jpg	1639-1311-697	2802-2606-1243	500-500-251	1196-1105-586

The next step is to eiffel-1.jpg,eiffel-13.jpg the two images by matching the results with different algorithms.
Sift

SURF

Orb

FAST

Places to be aware of:
* When linking, add a pkg-config opencv --cflags --libs library that can be added to all OPENCV
* Sift,surf is nonfree, need the method when use initModule_nonfree() , need header file opencv2/nonfree/nonfree.hpp
* Fast only detects corner points, to be combined with other extractor such as orb,sift.

Reference:
1.OPENCV feature2d
2.Feature Detection and Description

Comparison of Sift,surf,orb,fast feature descriptor operators in OpenCV