Digital processing method and system for determination of object occlusion in an image sequence

1. A method for occlusion region detection, comprising:

receiving, by a processing device, a first image and a second image;

estimating, by the processing device, a field of motion vectors between the first image and the second image;

motion compensating, by the processing device, the first image toward the second image to obtain a motion-compensated image;

comparing, by the processing device, a plurality of pixel values of the motion-compensated image to a plurality of pixels of the first image to estimate an error field;

inputting, by the processing device, the error field to a weighted error cost function to obtain an initial occlusion map; and

regularizing, by the processing device, the initial occlusion map to obtain a regularized occlusion map, wherein regularizing comprises applying a multi-sigma filter to the initial occlusion map.

2. The method of claim 1 , wherein regularizing further comprises obtaining a regularized error field.

3. The method of claim 2 , further comprising repeating said comparing and said regularizing until a value based on at least one of the regularized occlusion map or the regularized error field is below a threshold value.

4. The method of claim 1 , wherein motion compensating the first image toward the second image comprises image warping the field of motion vectors from the first image toward the second image.

5. The method of claim 1 , wherein the initial occlusion map and the regularized occlusion map are each based on a weighted error cost function.

6. The method of claim 5 , wherein the weighted error cost function is at least one of a sum-of-square differences measure, a locally scaled sum-of-square differences measure, a normalized cross-correlation measure, or a zero-mean normalized cross-correlation measure.

7. The method of claim 6 , wherein the weighted error cost function is based on a local weighting over a local region of support.

8. The method of claim 7 , wherein the local weighting over a local region of support is based on an eigensystem analysis of the local structure tensor of the motion-compensated image.

9. The method of claim 7 , wherein the local weighting over a local region of support is a gradient-energy weighting over the local region of support.

10. The method of claim 9 , wherein the gradient-energy weighting over a local region of support is a sum of statistical variance or local contrast over the local region of support.

11. The method of claim 1 , wherein applying a multi-sigma filter to the initial occlusion map comprises applying a 4 -factor sigma filter to the initial occlusion map.

12. The method of claim 11 , wherein input weights for the multi-factor sigma filter comprise an initial coarse occlusion field estimate and, between the first image and the second image one or more of similarities of color value or luminance, similarities of circular values of motion vector directions, or similarities of motion vector magnitudes.

13. The method of claim 12 , wherein the multi-factor sigma filter incorporates one or more weights, such as depth, or discontinuities of a range-to-target field.

14. A system, comprising:

a memory;

a processing device coupled to and having use of the memory, the processing device to:

receive a first image and a second image;

estimate a field of motion vectors between the first image and the second image;

motion compensate the first image toward the second image to obtain a motion-compensated image;

compare a plurality of pixel values of the motion-compensated image to a plurality of pixels of the first image to estimate an error field;

input the error field to a weighted error cost function to obtain an initial occlusion map; and

regularize the initial occlusion map to obtain a regularized occlusion map,

wherein regularizing comprises applying a multi-sigma filter to the initial occlusion map.

15. The system of claim 14 , wherein regularizing further comprises obtaining a regularized error field.

16. The system of claim 15 , further comprising repeating said comparing and said regularizing until a value based on at least one of the regularized occlusion map or the regularized error field is below a threshold value.

17. A non-transitory computer-readable storage medium including instructions that, when accessed by a processing device, cause the processing device to perform operations comprising:

receiving a first image and a second image;

estimating a field of motion vectors between the first image and the second image;

motion compensating the first image toward the second image to obtain a motion-compensated image;

comparing a plurality of pixel values of the motion-compensated image to a plurality of pixels of the first image to estimate an error field;

inputting the error field to a weighted error cost function to obtain an initial occlusion map; and

regularizing the initial occlusion map to obtain a regularized occlusion map, wherein regularizing comprises applying a multi-sigma filter to the initial occlusion map.

18. The non-transitory computer-readable storage medium of claim 17 , wherein regularizing further comprises obtaining a regularized error field.

19. The non-transitory computer-readable storage medium of claim 18 , further comprising repeating said comparing and said regularizing until a value based on at least one of the regularized occlusion map or the regularized error field is below a threshold value.

20. The non-transitory computer-readable storage medium of claim 17 , wherein the initial occlusion map and the regularized occlusion map are each based on a weighted error cost function.

21. The non-transitory computer-readable storage medium of claim 20 , wherein the weighted error cost function is based on a local weighting over a local region of support.

22. The non-transitory computer-readable storage medium of claim 21 , wherein the local weighting over a local region of support is based on an eigensystem analysis of the local structure tensor of the motion-compensated image.

23. The non-transitory computer-readable storage medium of claim 17 , wherein applying a multi-sigma filter to the initial occlusion map comprises applying a 4-factor sigma filter to the initial occlusion map.

24. The system of claim 14 , wherein applying a multi-sigma filter to the initial occlusion map comprises applying a 4-factor sigma filter to the initial occlusion map.