Method and apparatus for tracking superpixels between related images

1. A method for pixel mapping between an origin superpixel in a first image and a target superpixel in a second image, the method comprising:

determining feature vectors of origin pixels of the origin superpixel;

determining feature vectors of target pixels of the target superpixel; and

selecting a mapping pixel for an origin pixel among the target pixels based on a comparison of the feature vectors,

wherein a feature vector of a pixel comprises a relative distance of the pixel to a mass center of the superpixel and a topology value of the pixel.

2. The method according to claim 1 , wherein the first image and the second image are successive images of a sequence of images and/or multi-view images of a scene.

3. The method according to claim 1 , wherein the topology value of the pixel indicates a minimal distance of the pixel to a nearest superpixel border.

4. The method according to claim 1 , wherein the target pixels for an origin pixel are determined by:

determining a ray starting in a mass center of the origin superpixel and passing through the origin pixel;

determining a corresponding ray for the target superpixel starting in a mass center of the target superpixel; and

selecting target pixels along the corresponding ray starting in the mass center of the target superpixel.

5. The method according to claim 1 , wherein the mapping pixel for an origin pixel is selected by determining a best matching feature vector.

6. The method according to claim 5 , wherein a matching quality of a feature vector is determined using an Euclidian vector distance.

7. The method according to claim 6 , wherein the best matching feature vector is the feature vector resulting in a first found minimum of the Euclidian distance.

8. The method according to claim 4 , wherein for determining the corresponding ray for the target superpixel a scaling of the target superpixel relative to the origin superpixel is taken into account.

9. The method according to claim 8 , wherein the superpixel is divided into four quadrants and a horizontal and a vertical scaling is determined individually for each of the quadrants.

10. A computer readable non-transitory storage medium having stored therein instructions enabling pixel mapping between an origin superpixel in a first image and a target superpixel in a second image, wherein the instructions, when executed by a computer, cause the computer to:

determine feature vectors of origin pixels of the origin superpixel;

determine feature vectors of target pixels of the target superpixel; and

select a mapping pixel for an origin pixel among the target pixels based on a comparison of the feature vectors,

wherein a feature vector of a pixel comprises a relative distance of the pixel to a mass center of the superpixel and a topology value of the pixel.

11. The computer readable non-transitory storage medium according to claim 10 , wherein the first image and the second image are successive images of a sequence of images and/or multi-view images of a scene.

12. The computer readable non-transitory storage medium according to claim 10 , wherein the topology value of the pixel indicates a minimal distance of the pixel to a nearest superpixel border.

13. The computer readable non-transitory storage medium according to claim 10 , wherein the instructions cause the computer to determine the target pixels for an origin pixel by:

determining a ray starting in a mass center of the origin superpixel and passing through the origin pixel;

determining a corresponding ray for the target superpixel starting in a mass center of the target superpixel; and

selecting target pixels along the corresponding ray starting in the mass center of the target superpixel.

14. The computer readable non-transitory storage medium according to claim 10 , wherein the instructions cause the computer to select the mapping pixel for an origin pixel by determining a best matching feature vector.

15. The computer readable non-transitory storage medium according to claim 14 , wherein the instructions cause the computer to determine a matching quality of a feature vector using an Euclidian vector distance.

16. The computer readable non-transitory storage medium according to claim 15 , wherein the best matching feature vector is the feature vector resulting in a first found minimum of the Euclidian distance.

17. The computer readable non-transitory storage medium according to claim 13 , wherein the instructions cause the computer to take a scaling of the target superpixel relative to the origin superpixel into account for determining the corresponding ray for the target superpixel.

18. The computer readable non-transitory storage medium according to claim 17 , wherein the instructions cause the computer to divide the superpixel into four quadrants and to determine a horizontal and a vertical scaling individually for each of the quadrants.

19. An apparatus configured to perform pixel mapping between an origin superpixel in a first image and a target superpixel in a second image, the apparatus comprising:

a feature vector determining unit configured to determine feature vectors of origin pixels of the origin superpixel and to determine feature vectors of target pixels of the target superpixel; and

a mapping pixel selector configured to select a mapping pixel for an origin pixel among the target pixels based on a comparison of the feature vectors,

wherein a feature vector of a pixel comprises a relative distance of the pixel to a mass center of the superpixel and a topology value of the pixel.

20. The apparatus according to claim 19 , wherein the first image and the second image are successive images of a sequence of images and/or multi-view images of a scene.

21. The apparatus according to claim 19 , wherein the topology value of the pixel indicates a minimal distance of the pixel to a nearest superpixel border.

22. The apparatus according to claim 19 , wherein the apparatus is configured to determine the target pixels for an origin pixel by:

determining a ray starting in a mass center of the origin superpixel and passing through the origin pixel;

determining a corresponding ray for the target superpixel starting in a mass center of the target superpixel; and

selecting target pixels along the corresponding ray starting in the mass center of the target superpixel.

23. The apparatus according to claim 19 , wherein the mapping pixel selector is configured to select the mapping pixel for an origin pixel by determining a best matching feature vector.

24. The apparatus according to claim 23 , wherein the feature vector determining unit is configured to determine a matching quality of a feature vector using an Euclidian vector distance.

25. The apparatus according to claim 24 , wherein the best matching feature vector is the feature vector resulting in a first found minimum of the Euclidian distance.

26. The apparatus according to claim 22 , wherein the apparatus is configured to take a scaling of the target superpixel relative to the origin superpixel into account for determining the corresponding ray for the target superpixel.

27. The apparatus according to claim 26 , wherein the apparatus is configured to divide the superpixel into four quadrants and to determine a horizontal and a vertical scaling individually for each of the quadrants.

28. An apparatus configured to perform pixel mapping between an origin superpixel in a first image and a target superpixel in a second image, the apparatus comprising a processing device and a memory device having stored therein instructions, which, when executed by the processing device, cause the apparatus to:

determine feature vectors of origin pixels of the origin superpixel;

determine feature vectors of target pixels of the target superpixel; and

select a mapping pixel for an origin pixel among the target pixels based on a comparison of the feature vectors,

wherein a feature vector of a pixel comprises a relative distance of the pixel to a mass center of the superpixel and a topology value of the pixel.

29. The apparatus according to claim 28 , wherein the first image and the second image are successive images of a sequence of images and/or multi-view images of a scene.

30. The apparatus according to claim 28 , wherein the topology value of the pixel indicates a minimal distance of the pixel to a nearest superpixel border.

31. The apparatus according to claim 28 , wherein the instructions cause the apparatus to determine the target pixels for an origin pixel by:

determining a ray starting in a mass center of the origin superpixel and passing through the origin pixel;

determining a corresponding ray for the target superpixel starting in a mass center of the target superpixel; and

selecting target pixels along the corresponding ray starting in the mass center of the target superpixel.

32. The apparatus according to claim 28 , wherein the instructions cause the apparatus to select the mapping pixel for an origin pixel by determining a best matching feature vector.

33. The apparatus according to claim 32 , wherein the instructions cause the apparatus to determine a matching quality of a feature vector using an Euclidian vector distance.

34. The apparatus according to claim 33 , wherein the best matching feature vector is the feature vector resulting in a first found minimum of the Euclidian distance.

35. The apparatus according to claim 31 , wherein the instructions cause the apparatus to take a scaling of the target superpixel relative to the origin superpixel into account for determining the corresponding ray for the target superpixel.

36. The apparatus according to claim 35 , wherein the instructions cause the apparatus to divide the superpixel into four quadrants and to determine a horizontal and a vertical scaling individually for each of the quadrants.