1. SURF: Speeded-Up Robust Features
Advisor : Sheng-Jyh Wang
Student : 劉彥廷
2011/10/17
Computer Vision and Image Understanding (CVIU)

2. Outline Introduction Related Works Speed-Up Robust Features Detection
Description
Experiments
Conclusion

3. Outline Introduction Related Works Speed-Up Robust Features Detection
Description
Experiments
Conclusion

4. Introduction Why do we care about feature matching? Object Recognition
Wide baseline matching
Tracking

5. Challenges Types of variance Illumination Scale Rotation Affine
Perspective
We want to find
Repeatability、Distinctiveness features

6. Outline Introduction Related Works Speed-Up Robust Features Detection
Description
Experiments
Conclusion

7. Related Works Harris Corner Detector - Harris 1988
Laplacian of Gaussian - Lindeberg 1998
Difference of Gaussian - Lowe 2004

8. Related Works flat edge corner Illumination invariance !!!
Harris Corner Detector - Harris 1988
Laplacian of Gaussian - Lindeberg 1998
Difference of Gaussian - Lowe 2004
flat
edge
corner
Illumination invariance !!!

9. * Related Works = Harris Corner Detector - Harris 1988
Laplacian of Gaussian - Lindeberg 1998
Difference of Gaussian - Lowe 2004
characteristic scale * =
LoG can detect blob-like structures at locations
“Feature Detection with Automatic Scale Selection”, IJCV ‘98

10. Related Works Harris Corner Detector - Harris 1988
Laplacian of Gaussian - Lindeberg 1998
Difference of Gaussian - Lowe 2004
Computational efficiency !
Compare to 26 neighbors
Keep the same keypoint in all scale !

11. Motivation
Lindeberg uses Laplacian of Gaussian, one could obtain scale invariant features.
Lowe uses difference of Gaussian to approximate Laplacian of Gaussian. (SIFT)
This paper uses Hessian - Laplacian to approximate Laplacian of Gaussian, to improve calculation speed.

12. Outline Introduction Related Works Speed-Up Robust Features Detection
Description
Experiments
Conclusion

13. Detection Hessian-based interest point localization
Lxx(x,y,σ) is the Laplacian of Gaussian of the image.
It is the convolution of the Gaussian second order derivative with the image.
This paper use Dxx to approximate Lxx.

14. Detection Scale analysis with constant image size
Approximated second order derivatives with box filters.
(DoG)
但是拿Gaussian derivatives來做，需要對Gaussian Function做離散化和裁切 比較浪費時間，尤其是filter大小又需要改變的時候

15. Integral Images Using integral images for major speed up
Integral Image (summed area tables) is an intermediate representation for the image and contains the sum of gray scale pixel values of image.
They can be evaluated at a very low computational cost using integral images
ABCD所圍面積內intensity的總和 = A-B-C+D
They can be evaluated at a very low computational cost using integral images with box filters

16. Summary
Keypoint detection
Keypoint description
Keypointmatching

17. Fourier v.s. Wavelet Fourier Transform (FT) is not a good tool –
gives no direct information about when an oscillation occurred.
Wavelets can keep track of time and frequency information.
Fourier basis
Haar basis

18. Description interest point Haar dx scale = s r = 6s dy
Orientation Assignment
The Haar wavelet responses are represented as vectors
Sum all responses within a sliding orientation window covering an angle of 60 degree
The longest vector is the dominant orientation
x response
y response dx dy
r = 6s
interest point
scale = s
Haar

19. Description
Split the interest region (20s x 20s) up into 4 x 4 square sub-regions.
Calculate Haar wavelet response dx and dy and weight the response with a Gaussian kernel.
Sum the response over each sub-region for dx and dy, then sum the absolute value of resp- onse.

20. Matching  can do match  can do match  not match 20
Fast indexing through the sign of the Laplacian for the underlying interest point
The sign of trace of the Hessian matrix
Trace = Lxx + Lyy
 can do match
 can do match
 not match
matching 20

21. Outline Introduction Related Works Speed-Up Robust Features Detection
Description
Experiments
Conclusion

22. Experiments Test keypoint repeatability for
(Viewpoint Change), (Lighting Change) and(Zoom and Rotation)

23. Experiments
Repeatability score for image sequences

24. Experiments
Fix number of keypoints

25. Experiments
SIFT
SURF
Leila Mirmohamadsadeghi , “Image Tag Propagation “ ‘10

26. Experiments
Image size : 341x341
Running time : seconds

27. Experiments
Image size : 800x600
Running time : seconds

28. Conclusion
SURF is faster than SIFT by 3 times, and has recall precision not worse than SIFT.
SURF is good at handling image with blurring or rotation.
SURF is poor at handling image with viewpoint .

29. Reference “Speeded-Up Robust Features”, CVIU ‘08 Herbert Bay
“Distinctive Image Features from Scale-Invariant Features”, IJCV ’04 David G. Lowe
“A Combined Corner and Edge Detector” ‘88 Chris Harris
“Feature Detection with Automatic Scale Selection”, IJCV ’98 Lindeberg