Method and apparatus for tracking features in a video sequence

• US 20040120548A1
• Filed: 12/18/2002
• Published: 06/24/2004
• Est. Priority Date: 12/18/2002
• Status: Active Grant

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Qian, Richard J.
• Granted Patent: US 7,194,110 B2
• Time in Patent Office: Days
• US Class Current: 382/103
• CPC Class Codes: G06T 7/246 using feature-based methods, e.g. the tracking of corners or segments

Abstract

A method for performing feature tracking includes generating a pool of candidates on an image frame for a first feature point from information about a pool of candidates for a second feature point on a previous image frame.

Claims

1. A method for performing feature tracking on image frames, comprising:

initializing feature points on a first image frame;

generating candidates for each of the feature points on a second image frame; and

generating candidates for a first feature point on a third image frame from information about candidates for a second feature point on the second image frame

2. The method of claim 1, wherein initializing the feature points on the first image frame comprises utilizing user responses.

3. The method of claim 1, wherein initializing the feature points on the first image frame comprises performing feature point detection.

4. The method of claim 1, wherein generating the candidates comprises performing template matching.

5. The method of claim 1, wherein generating the candidates comprises performing feature point matching.

6. The method of claim 1, wherein generating candidates for the first feature point on the third image frame from the information about the candidates for the second feature point on the second image frame comprises:

selecting one of the candidates for the second feature point on the second image frame;

projecting a fixed distance between the first and second feature points from a position of the candidate selected; and

selecting a candidate in a neighborhood of the fixed distance.

7. The method of claim 1, further comprising generating candidates for the first feature point on the third image frame from information about the candidates for a third feature point on the second image frame.

8. The method of claim 1, wherein generating candidates for the first feature point on the third image frame comprises generating a series of temporary candidates based on information about previously generated candidates.

9. The method of claim 1, further comprising generating probability values for each of the candidates for the first feature point on the third image frame.

10. The method of claim 9, wherein generating probability values for each of the candidates for the first feature point on the third image frame comprises comparing each of the candidates with a selected candidate for the first feature point on the second image frame.

11. The method of claim 9, wherein generating probability values for each of the candidates for the first feature point on the third image frame comprises comparing each of the candidates with an average of the candidates for the first feature point on the second image frame.

12. The method of claim 9, wherein generating probability values for each of the candidates for the first feature point on the third image frame comprises comparing each of the candidates with an average of selected candidate for the first feature point for previous image frames.

13. The method of claim 1, further comprising selecting a candidate for each of the feature points on the second image frame.

14. The method of claim 1, further comprising selecting a candidate for the first feature point on the third image frame.

15. A method for performing feature tracking, comprising:

generating a pool of candidates on an image frame for a first feature point from information about a pool of candidates for a second feature point on a previous image frame.

16. The method of claim 15, wherein generating the pool of candidates on the image frame for the first feature point from information about the pool of candidates for the second feature point on the previous image frame comprises:

selecting one of the candidates from the pool of candidates for the second feature point;

projecting a fixed distance between the first and second feature points from a position of a candidate selected; and

selecting a candidate from a neighborhood of the fixed distance.

17. The method of claim 15, wherein generating the pool of candidates on the image frame for the first feature point from information about the pool of candidates for the second feature point on the previous image frame comprises:

generating a series of pools of temporary candidates for the first feature point from a series of pools of pre-existing temporary candidates associated with the second feature point; and

designating a last pool in the series of pools of temporary candidates for the first feature point as the pool of candidates for the first feature point.

18. The method of claim 15, wherein generating a pool of candidates for a first feature point on a image frame from a pools of candidates for a second feature point on a previous image frame comprises utilizing sequential Monte Carlo sampling.

19. The method of claim 15, further comprising generating probabilities for the pool of candidates for the first feature.

20. A machine-readable medium having stored thereon sequences of instructions, the sequences of instructions including instructions which, when executed by a processor, causes the processor to perform:

initializing feature points on a first image frame;

generating candidates for each of the feature points on a second image frame; and

generating candidates for a first feature point on a third image frame from information about candidates for a second feature point on the second image frame.

21. The machine-readable medium of claim 20, wherein generating candidates for the first feature point on the third image frame from the information about the candidates for the second feature point on the second image frame comprises:

selecting one of the candidates for the second feature point on the second image frame;

projecting a fixed distance between the first and second feature points from a position of the candidate selected; and

selecting a candidate in a neighborhood of the fixed distance.

22. The machine-readable medium of claim 20, further comprising instructions which, when executed by the processor, causes the processor to perform generating candidates for the first feature point on the third image frame from information about the candidates for a third feature point on the second image frame.

23. The machine-readable medium of claim 20, wherein generating candidates for the first feature point on the third image frame comprises generating a series of temporary candidates based on information about previously generated candidates.

24. The machine-readable medium of claim 20, further comprising instructions which, when executed by the processor, causes the processor to perform generating probability values for each of the candidates for the first feature point on the third image frame.

25. An feature tracking unit, comprising:

a candidate pool evolution unit to generate a pool of candidates for a first feature point on an image frame from information from a pool of candidates for a second feature point on a previous image frame.

26. The apparatus of claim 25, further comprising a feature point matching unit to generate probability values for the pool of candidates for the first feature point.

27. The apparatus of claim 25, further comprising a feature point matching unit to select a candidate from the pool of candidates for the first feature point as an observed first feature point for the image frame.

28. The apparatus of claim 25, further comprising a rotation and translation estimation unit to estimate a change in position of the first feature point at the image frame.