Corrective lighting for video inpainting
One or more processing devices access a scene depicting a reference object that includes an annotation identifying a target region to be modified in one or more video frames. The one or more processing devices determine that a target pixel corresponds to a sub-region within the target region that includes hallucinated content. The one or more processing devices determine gradient constraints using gradient values of neighboring pixels in the hallucinated content, the neighboring pixels being adjacent to the target pixel and corresponding to four cardinal directions. The one or more processing devices update color data of the target pixel subject to the determined gradient constraints.
1. A computer implemented method in which one or more processing devices perform operations comprising:
identifying a target region to be modified in one or more video frames;
determining that a target pixel is in a sub-region of the target region;
based on determining that the target pixel is in the sub-region, hallucinating content into the sub-region;
after hallucinating the content into the sub-region, identifying neighboring pixels that are adjacent to the target pixel in the sub-region in four cardinal directions, the neighboring pixels being part of the hallucinated content;
determining gradient constraints using gradient values of the identified neighboring pixels in the hallucinated content; and
updating color data of the target pixel subject to the determined gradient constraints.
2. The computer implemented method of claim 1 , wherein determining the gradient constraints comprises computing a forward-flow-traced gradient constraint in the four cardinal directions.
3. The computer implemented method of claim 1 , wherein determining the gradient constraints comprises computing a backward-flow-traced gradient constraint in the four cardinal directions.
4. The computer implemented method of claim 1 , further comprising:
determining that a target motion of the target pixel cannot be computed subject to a motion constraint; and
based on determining that the target motion of the target pixel cannot be computing subject to the motion constraint, hallucinating the content into the sub-region.
5. The computer implemented method of claim 1 , further comprising:
identifying an area of the target region for which sparse feature points cannot be determined; and
selecting the identified area as the sub-region.
6. The computer implemented method of claim 1 , wherein the sub-region is a portion of the target region that does not include a forward-flow-traced position or a backward-flow traced position.
7. The computer implemented method of claim 1 , wherein the content is hallucinated into the sub-region by video inpainting using a reference frame.
8. The computer implemented method of claim 1 , further comprising applying a weighting function to the gradient constraints, wherein the weighting function prioritizes information from the hallucinated content.
9. A computing system comprising;
a processing device; and
a non-transitory computer-readable medium communicatively coupled to the processing device and storing program code, the processing device configured to execute the program code and thereby performing operations comprising:
identifying a target region to be modified in one or more video frames;
determining that a target pixel is in a sub-region of the target region;
based on determining that the target pixel is in the sub-region, hallucinating content into the sub-region;
after hallucinating the content into the sub-region, identifying neighboring pixels that are adjacent to the target pixel in the sub-region in four cardinal directions, the neighboring pixels being part of the hallucinated content;
determining gradient constraints using gradient values of the identified neighboring pixels in the hallucinated content; and
updating color data of the target pixel subject to the determined gradient constraints.
10. The computing system of claim 9 , the operations further comprising:
determining that a target motion of the target pixel cannot be computed subject to a motion constraint; and
based on determining that the target motion of the target pixel cannot be computing subject to the motion constraint, hallucinating the content into the sub-region.
11. The computing system of claim 9 , the operations further comprising:
identifying an area of the target region for which sparse feature points cannot be determined; and
selecting the identified area as the sub-region.
12. The computing system of claim 9 , wherein the sub-region is a portion of the target region that does not include a forward-flow-traced position or a backward-flow traced position.
13. The computing system of claim 9 , wherein the content is hallucinated into the sub-region by video inpainting using a reference frame.
14. The computing system of claim 9 , the operations further comprising applying a weighting function to the gradient constraints, wherein the weighting function prioritizes information from the hallucinated content.
15. A non-transitory computer-readable medium having program code of a video editing tool stored thereon, wherein the program code, when executed by one or more processing devices, configures the one or more processing devices to perform operations comprising:
identifying a target region to be modified in one or more video frames;
determining that a target pixel is in a sub-region of the target region;
based on determining that the target pixel is in the sub-region, hallucinating content into the sub-region;
after hallucinating the content into the sub-region, identifying neighboring pixels that are adjacent to the target pixel in the sub-region in four cardinal directions, the neighboring pixels being part of the hallucinated content;
determining gradient constraints using gradient values of the identified neighboring pixels in the hallucinated content; and
updating color data of the target pixel subject to the determined gradient constraints.
16. The non-transitory computer-readable medium of claim 15 , the operations further comprising:
determining that a target motion of the target pixel cannot be computed subject to a motion constraint; and
based on determining that the target motion of the target pixel cannot be computing subject to the motion constraint, hallucinating the content into the sub-region.
17. The non-transitory computer-readable medium of claim 15 , the operations further comprising:
identifying an area of the target region for which sparse feature points cannot be determined; and
selecting the identified area as the sub-region.
18. The non-transitory computer-readable medium of claim 15 , wherein the sub-region is a portion of the target region that does not include a forward-flow-traced position or a backward-flow traced position.
19. The non-transitory computer-readable medium of claim 15 , wherein the content is hallucinated into the sub-region by video inpainting using a reference frame.
20. The non-transitory computer-readable medium of claim 15 , the operations further comprising applying a weighting function to the gradient constraints, wherein the weighting function prioritizes information from the hallucinated content.