Systems and methods for digital video stabilization via constraint-based rotation smoothing
Systems and methods for digital video stabilization via constraint-based rotation smoothing are provided. Digital video data including a set of image frames having associated time stamps and a set of camera orientation data having associated time stamps may be provided. A smoothed set of camera orientation data may be generated by minimizing a rate of rotation between successive image frames while minimizing an amount of empty regions in a resulting set of smoothed image frames reoriented based on the smoothed set of camera orientation data.
1. A computer-implemented method comprising:
acquiring, by a computing system, video data comprising a set of image frames captured by a recording device;
acquiring, by the computing system, gyroscope trace data associated with the video data; and
estimating, by the computing system, one or more video capture characteristics of the recording device based on the video data and the gyroscope trace data, wherein
the one or more video capture characteristics of the recording device include at least one of: a focal length, a rolling shutter duration, a gyroscope delay, or a gyroscope drift of the recording device.
2. The computer-implemented method of claim 1 further comprising identifying one or more matching points in one or more sets of consecutive image frames of the set of image frames, the one or more matching points defining a set of point correspondences, wherein
the estimating one or more video capture characteristics of the recording device comprises estimating one or more video capture characteristics of the recording device based on the set of point correspondences.
3. The computer implemented method of claim 2 , wherein the estimating one or more video capture characteristics of the recording device based on the set of point correspondences comprises:
minimizing a mean-square re-projection error of all point correspondences in the set of point correspondences.
4. The computer-implemented method of claim 3 , wherein the minimizing the mean-square re-projection error of all point correspondences in the set of point correspondences is performed using a non-linear optimizer.
5. The computer-implemented method of claim 2 , wherein the one or more matching points are identified based on a scale invariant feature transform.
6. The computer-implemented method of claim 2 , wherein the identifying one or more matching points in one or more sets of consecutive image frames of the set of image frames further comprises discarding outlier data using random sample consensus.
7. The computer-implemented method of claim 1 , wherein the video data comprises a video capture of a stationary object captured while the recording device is being shaken.
8. The computer-implemented method of claim 1 , wherein the video data comprises a video capture of ten seconds or less.
9. The computer-implemented method of claim 1 , wherein the recording device is a rolling shutter recording device.
10. A system comprising:
at least one processor; and
a memory storing instructions that, when executed by the at least one processor, cause the system to perform a method comprising:
acquiring video data comprising a set of image frames captured by a recording device;
acquiring, by the computing system, gyroscope trace data associated with the video data; and
estimating one or more video capture characteristics of the recording device based on the video data and the gyroscope trace data, wherein
the one or more video capture characteristics of the recording device include at least one of: a focal length, a rolling shutter duration, a gyroscope delay, or a gyroscope drift of the recording device.
11. The system of claim 10 , wherein the method further comprises identifying one or more matching points in one or more sets of consecutive image frames of the set of image frames, the one or more matching points defining a set of point correspondences, and further wherein
the estimating one or more video capture characteristics of the recording device comprises estimating one or more video capture characteristics of the recording device based on the set of point correspondences.
12. The system of claim 11 , wherein the estimating one or more video capture characteristics of the recording device based on the set of point correspondences comprises:
minimizing a mean-square re-projection error of all point correspondences in the set of point correspondences.
13. The system of claim 12 , wherein the minimizing the mean-square re−projection error of all point correspondences in the set of point correspondences is performed using a non-linear optimizer.
14. A non-transitory computer-readable storage medium including instructions that, when executed by at least one processor of a computing system, cause the computing system to perform a method comprising:
acquiring video data comprising a set of image frames captured by a recording device;
acquiring, by the computing system, gyroscope trace data associated with the video data; and
estimating one or more video capture characteristics of the recording device based on the video data and the gyroscope trace data, wherein
the one or more video capture characteristics of the recording device include at least one of: a focal length, a rolling shutter duration, a gyroscope delay, or a gyroscope drift of the recording device.
15. The non-transitory computer-readable storage medium of claim 14 , wherein the method further comprises identifying one or more matching points in one or more sets of consecutive image frames of the set of image frames, the one or more matching points defining a set of point correspondences, and further wherein
the estimating one or more video capture characteristics of the recording device comprises estimating one or more video capture characteristics of the recording device based on the set of point correspondences.
16. The non-transitory computer-readable storage medium of claim 15 , wherein the estimating one or more video capture characteristics of the recording device based on the set of point correspondences comprises:
minimizing a mean-square re-projection error of all point correspondences in the set of point correspondences.
17. The non-transitory computer-readable storage medium of claim 16 , wherein the minimizing the mean-square re-projection error of all point correspondences in the set of point correspondences is performed using a non-linear optimizer.