Claims
- 1. A method for estimating motion of a first region from an original video frame to a target video frame, the method comprising:
calculating error map for the first region; calculating a plurality of error maps for neighboring regions around the first region; biasing the error map for the first region using the error maps for the neighboring regions; and determining the estimated motion for the first region using the biased error map.
- 2. The method of claim 1, wherein the neighbor regions are adjacent to the first region in the original video frame.
- 3. The method of claim 1, wherein the error maps are calculated using sums of absolute differences or L1 error.
- 4. The method of claim 1, wherein the error maps are calculated using an Lp norm where p is a positive real number.
- 5. The method of claim 1, wherein the error maps are calculated using any aggregate statistical scalar function as an error measure.
- 6. The method of claim 1, further comprising:
weighting the error maps for the neighboring regions relative to each other prior to performing the biasing.
- 7. The method of claim 6, further comprising:
calculating a composite weighted error map from the weighted error maps for the neighboring regions; and using the composite weighted error map in biasing the error map for the first region.
- 8. The method of claim 7, wherein relative motion between each of the neighboring regions and the first region contributes to the weighting.
- 9. The method of claim 7, wherein proximity of each of the neighbor regions to the first region contributes to the weighting.
- 10. The method of claim 9, wherein the proximity comprises an amount of overlap between the neighbor region and an annulus-like area surrounding the first region.
- 11. The method of claim 9, wherein the proximity comprises an amount of overlap between the neighbor region and a bloated bounding box surrounding the first region.
- 12. The method of claim 7, wherein a color similarity between each of the neighboring regions and the first region contributes to the weighting.
- 13. The method of claim 7, wherein a relative size between each of the neighboring regions and the first region contributes to the weighting.
- 14. The method of claim 7, wherein the weighting comprises a product of multiple contributing factors, and wherein the multiple contributing factors include a relative motion measure, a proximity measure, and a color similarity measure.
- 15. The method of claim 1, wherein the first region and the neighboring regions comprise segments obtained by an image segmentation process.
- 16. The method of claim 1, wherein the first region and the neighboring regions comprise arbitrarily-shaped regions.
- 17. The method of claim 1, wherein the source frame and the target frame comprise video frames from a same group of pictures (GOP).
- 18. The method of claim 1, further comprising:
determining the estimated motion by finding a local minimum in the biased error map for the first region.
- 19. The method of claim 18, wherein the local minimum is required to be below a threshold value in order to determine an acceptable estimation.
- 20. An apparatus for estimating motion of a first region from an original video frame to a target video frame, the apparatus comprising:
means for calculating error map for the first region; means for calculating a plurality of error maps for neighboring regions around the first region; means for biasing the error map for the first region using the error maps for the neighboring regions; and means for determining the estimated motion for the first region using the biased error map.
- 21. The apparatus of claim 20, wherein the apparatus comprises an encoder.
- 22. The apparatus of claim 21, wherein the apparatus comprises a decoder configured to receive video information from the encoder, and wherein the decoder also includes the means for calculating error map for the first region, the means for calculating a plurality of error maps for neighboring regions around the first region, the means for biasing the error map for the first region using the error maps for the neighboring regions, and the means for determining the estimated motion for the first region using the biased error map.
- 23. A method for estimating motion of a first region from an original video frame to a target video frame, the method comprising:
calculating an unbiased error map for the first region; determining an unconstrained estimated motion for the first region using the unbiased error map; calculating error values in the vicinity of the unconstrained estimated motion for a plurality of neighboring regions around the first region; and adjusting the unconstrained estimated motion using the calculated error values to determine a locally-constrained estimated motion for the first region.
CROSS-REFERENCES TO RELATED APPLICATIONS
[0001] This patent application claims priority from U.S. Provisional Patent Application No. 60/274,079, filed Mar. 7, 2001, and entitled “Motion Matching Using Local Segment Motion Constraints,” the disclosure of which is incorporated herein by reference. This patent application is related to U.S. patent application Ser. No. 09/550,705, filed Apr. 17, 2000 and entitled “Method and Apparatus for Efficient Video Processing,” the disclosure of which is incorporated herein by reference. This invention is also related to U.S. patent application Ser. No. 09/591,438, filed Jun. 9, 2000 and entitled “Method and Apparatus for Digital Image Segmentation,” the disclosure of which is incorporated herein by reference.
Provisional Applications (1)
|
Number |
Date |
Country |
|
60274079 |
Mar 2001 |
US |