Depth mask assisted video stabilization

1. A method for depth mask assisted video stabilization comprising:

creating a first depth mask from a first frame and a second depth mask from a second frame to obtain depth information;

defining a first list of feature points from the first frame and a second list of feature points from the second frame;

comparing feature points from the first list of feature points with feature points from the second list of feature points to find a number of common feature point pairs;

creating a number of motion matrices from the common feature point pairs;

including the motion matrices in a correction matrix; and

applying the inverse of the correction matrix to the second frame to perform video stabilization.

2. The method of claim 1 , wherein defining the first list of feature points includes grouping pixels that are at a common depth and qualifying feature points based on contrast, uniqueness, and spatial location.

3. The method of claim 1 , wherein comparing feature points includes comparing feature points that are at a common depth.

4. The method of claim 1 , wherein including the motion matrices in a correction matrix includes including those motion matrices in the correction matrix that do not have subject motion where subject motion is determined by a particular motion matrix that differs from a majority of the motion matrices.

5. The method of claim 4 , wherein creating a number of motion matrices from the common feature point pairs includes performing the best fit operation on the common feature point pairs to create the motion matrices.

6. A system for depth mask assisted video stabilization, comprising:

a memory;

a processor coupled to the memory to:

create a first depth mask from a first frame and a second depth mask from a second frame to obtain depth information;

define a first list of depth objects from the first frame and a second list of depth objects from the second frame;

compare depth objects from the first list of depth objects with depth objects from the second list of depth objects to find a number of common depth object pairs;

perform a best fit operation on the common depth object pairs to create a number of motion matrices;

include the motion matrices in a correction matrix; and

apply the inverse of the correction matrix to the second frame to perform video stabilization.

7. The system of claim 6 , wherein pixels at a common depth are grouped together to define depth objects.

8. The system of claim 7 , wherein a pixel's depth is defined by reference to the corresponding depth mask.

9. The system of claim 6 , wherein depth objects from the first list of depth objects and depth objects from the second list of depth objects are compared if the depth objects are at a common depth.

10. The system of claim 6 , wherein the motion matrices that contain subject motion are removed from the correction matrix before the correction matrix is applied to the second frame where a motion matrix contains subject motion when the motion matrix differs from the majority of the motion matrices.

11. A non-transitory computer-readable medium storing instructions for depth mask assisted video stabilization executable by a computer to cause a computer to:

create a first depth mask from a first frame and a second depth mask from a second frame to obtain depth information;

define a first list of feature points from the pixels in the first frame and a second list of feature points from the pixels in the second frame;

compare feature points from the first list of feature points with feature points from the second list of feature points to find a number of common feature point pairs;

create a number of common depth objects from the common feature point pairs;

perform a best fit operation on the common depth objects to create a number of motion matrices;

include the motion matrices that do not contain subject motion in a correction matrix where a motion matrix contains subject motion when the motion matrix differs from the majority of the motion matrices; and

apply the inverse of the correction matrix to the second frame to perform video stabilization.

12. The medium of claim 11 , wherein the best fit operation that is performed on the common depth objects includes performing the best fit operation on the common feature point pairs that compose a common depth object.

13. The medium of claim 11 , wherein feature points from the first list of feature points and feature points from the second list of feature points are compared if the feature points are at a common depth.

14. The medium of claim 11 , wherein a common depth object is created by grouping a number of common feature point pairs.

15. The medium of claim 14 , wherein the common feature point pairs are grouped together based on depth and spatial location within the first and second frame.