Blue noise mask for video sampling

1. A method comprising, by one or more computing systems:

receiving a video comprising a sequence of frames;

accessing a three-dimensional mask that specifies pixel-sampling locations, the three-dimensional mask having (1) a first dimension and a second dimension corresponding to a spatial domain and (2) a third dimension corresponding to a temporal domain, wherein a blue noise property is present in:

the pixel-sampling locations that are associated with each of a plurality of two-dimensional spatial slices of the three-dimensional mask in the spatial domain; and

the pixel-sampling locations that are associated with each of a plurality of one-dimensional temporal slices of the three-dimensional mask in the temporal domain;

generating a sample of the video by sampling the sequence of frames using the three-dimensional mask; and

generating a reconstructed video from the generated sample of the video using a machine-learning model.

2. The method of claim 1 , wherein the three-dimensional mask is generated by selecting the pixel-sampling locations using a distance function, wherein the distance function measures a distance between any two pixel-sampling locations in the three-dimensional mask based on a Euclidean distance in the spatial domain and a Manhattan distance in the temporal domain.

3. The method of claim 2 , wherein the pixel-sampling locations specified by the three-dimensional mask are sequentially selected, wherein at least one of the pixel-sampling locations is selected based on distances, measured using the distance function, between that pixel-sampling location and previously selected pixel-sampling locations.

4. The method of claim 1 , wherein the pixel-sampling locations in the three-dimensional mask are associated with sequential values corresponding to a sequence in which the pixel-sampling locations are selected.

5. The method of claim 4 , further comprising:

determining a threshold value corresponding to a desired percentage of the pixel-sampling locations to be used for generating the sample of the video; and

selecting a subset of the pixel-sampling locations based on the threshold value and the sequential values associated with the pixel-sampling locations;

wherein the sample of the video is generated by sampling the sequence of frames according to the subset of the pixel-sampling locations.

6. The method of claim 1 , wherein the sample of the video has incomplete pixel information and the reconstructed video has complete pixel information.

7. The method of claim 1 , further comprising:

transmitting the sample of the video to a second computing system to reconstruct the video at the second computing system.

8. One or more computer-readable non-transitory storage media embodying software that is operable when executed to:

receive a video comprising a sequence of frames;

access a three-dimensional mask that specifies pixel-sampling locations, the three-dimensional mask having (1) a first dimension and a second dimension corresponding to a spatial domain and (2) a third dimension corresponding to a temporal domain, wherein a blue noise property is present in:

the pixel-sampling locations that are associated with each of a plurality of two-dimensional spatial slices of the three-dimensional mask in the spatial domain; and

the pixel-sampling locations that are associated with each of a plurality of one-dimensional temporal slices of the three-dimensional mask in the temporal domain;

generate a sample of the video by sampling the sequence of frames using the three-dimensional mask; and

generate a reconstructed video from the generated sample of the video using a machine-learning model.

9. The media of claim 8 , wherein the three-dimensional mask is generated by selecting the pixel-sampling locations using a distance function, wherein the distance function measures a distance between any two pixel-sampling locations in the three-dimensional mask based on a Euclidean distance in the spatial domain and a Manhattan distance in the temporal domain.

10. The media of claim 9 , wherein the pixel-sampling locations specified by the three-dimensional mask are sequentially selected, wherein at least one of the pixel-sampling locations is selected based on distances, measured using the distance function, between that pixel-sampling location and previously selected pixel-sampling locations.

11. The media of claim 8 , wherein the pixel-sampling locations in the three-dimensional mask are associated with sequential values corresponding to a sequence in which the pixel-sampling locations are selected.

12. The media of claim 11 , wherein the software is further operable when executed to:

determine a threshold value corresponding to a desired percentage of the pixel-sampling locations to be used for generating the sample of the video; and

select a subset of the pixel-sampling locations based on the threshold value and the sequential values associated with the pixel-sampling locations;

wherein the sample of the video is generated by sampling the sequence of frames according to the subset of the pixel-sampling locations.

13. The media of claim 8 , wherein the sample of the video has incomplete pixel information and the reconstructed video has complete pixel information.

14. The media of claim 8 , wherein the software is further operable when executed to:

transmit the sample of the video to a second computing system to reconstruct the video at the second computing system.

15. A system comprising:

one or more processors; and

one or more computer-readable non-transitory storage media coupled to one or more of the processors and comprising instructions operable when executed by one or more of the processors to cause the system to:

receive a video comprising a sequence of frames;

access a three-dimensional mask that specifies pixel-sampling locations, the three-dimensional mask having (1) a first dimension and a second dimension corresponding to a spatial domain and (2) a third dimension corresponding to a temporal domain, wherein a blue noise property is present in:

the pixel-sampling locations that are associated with each of a plurality of two-dimensional spatial slices of the three-dimensional mask in the spatial domain; and

the pixel-sampling locations that are associated with each of a plurality of one-dimensional temporal slices of the three-dimensional mask in the temporal domain;

generate a sample of the video by sampling the sequence of frames using the three-dimensional mask; and

generate a reconstructed video from the generated sample of the video using a machine-learning model.

16. The system of claim 15 , wherein the three-dimensional mask is generated by selecting the pixel-sampling locations using a distance function, wherein the distance function measures a distance between any two pixel-sampling locations in the three-dimensional mask based on a Euclidean distance in the spatial domain and a Manhattan distance in the temporal domain.

17. The system of claim 16 , wherein the pixel-sampling locations specified by the three-dimensional mask are sequentially selected, wherein at least one of the pixel-sampling locations is selected based on distances, measured using the distance function, between that pixel-sampling location and previously selected pixel-sampling locations.

18. The system of claim 15 , wherein the pixel-sampling locations in the three-dimensional mask are associated with sequential values corresponding to a sequence in which the pixel-sampling locations are selected.

19. The system of claim 18 , wherein the processors are further operable when executing the instructions to:

determine a threshold value corresponding to a desired percentage of the pixel-sampling locations to be used for generating the sample of the video; and

select a subset of the pixel-sampling locations based on the threshold value and the sequential values associated with the pixel-sampling locations;

wherein the sample of the video is generated by sampling the sequence of frames according to the subset of the pixel-sampling locations.

20. The system of claim 15 , wherein the processors are further operable when executing the instructions to:

transmit the sample of the video to a second computing system to reconstruct the video at the second computing system.