Presentation slides for talk given at Computer Robot Vision 2010 in Ottawa, Canada.

1. Global Context Descriptors for SURF and MSER Feature Descriptors Gail Carmichael Robert Laganière Prosenjit Bose

2. Gail Carmichael PhD Student, School of Computer Science http://www.gailcarmichael.com

3. Stuff I care about... Learning

4. Stuff I care about... Augmented Reality

5. Stuff I care about... Matching Panoramas and Photographs

6. SIFT with Global Context Mortensen, Deng, Shapiro (2005)

7. SIFT with Global Context Mortensen, Deng, Shapiro (2005)

8. SIFT with Global Context Mortensen, Deng, Shapiro (2005)

9. Global, But More Local Li and Ma (2009)

10. Global, But More Local Li and Ma (2009)

11. Global Context for SURF Pretty much the same as SIFT

12. Global Context for SURF 1. Find SURF points with SURFmex

13. Global Context for SURF 2. Find nearby features for each SURF point

14. Global Context for SURF 3. Put curvature values in log-polar graph.

15. Global Context for SURF 4. Append values of bins to get Global Context.

16. Global Context for MSER A little different from SIFT/SURF...

17. Global Context for MSER 1. Detect Maximally Stable Extremal Regions (MSERs) with VLFeat

18. Global Context for MSER 2a. Create shape patches (Forssén and Lowe).

19. Global Context for MSER 2b. Create texture patches (Forssénand Lowe).

20. Global Context for MSER Measurement region? Points to measure? Multiple of bounding ellipse. All values around MSER.

21. Global Context for MSER 3. Create a curvature value patch.

22. Global Context for MSER 4. Collect curvature values in log-polar graph,append values to become Global Context.

23. Matching Results

24. Matching Results

25. Matching Results

26. Matching Results

27. Matching Results

28. Matching Results

29. Matching Results

30. Matching Results

31. Matching Results

32. Future Work Elliptical neighbouring region for SURF Minimum measurement region for MSER More tests, more images

33. Thanks!