Constraint model space partition tree for video stabilization
AssigneeGoPro, Inc.
InventorsThomas Derbanne, Nicolas Rahmouni, Luis Mario Domenzain
Life of this patent
Estimating the term…
Abstract
A constraint model may include a representation of feasible viewing window placement within the field of view of a video for stabilization. When the viewing window extends beyond the field of view of the video, the placement of the viewing window may be modified using a constraint model space partition tree. The constraint model space partition tree may include branches that split the constraint model and enable use of greater number of stabilization parameters to stabilize the video.
Claims20
A system for stabilizing videos, the system comprising:
- one or more physical processors configured by machine-readable instructions to:
- obtain video information defining a video, the video including visual content having a progress length, the visual content having a source field of view;
- obtain viewing window placement information for the video, the viewing window placement information defining placement of a viewing window for the visual content as the function of progress through the progress length, the viewing window defining extents of the visual content to be included within stabilized visual content as the function of progress through the progress length, the viewing window having a punchout field of view;
- obtain constraint model space partition tree information for the video, the constraint model space partition tree information defining a constraint model space partition tree for the video, the constraint model space partition tree including branches that split a constraint model for the video, the constraint model for the video including a representation of feasible viewing window placement in which entirety of the punchout field of view of the viewing window is contained within the source field of view of the visual content, wherein the branches of the constraint model space partition tree terminate at leaves of the constraint model space partition tree, an individual leaf of the constraint model space partition tree defining a single projection function for a group of neighboring points inside the constraint model for the video to move the viewing window outside the source field of view of the visual content to inside the source field of view of the visual content;
- responsive to the placement of the viewing window outside the source field of view of the visual content, determine modified placement of the viewing window based on the constraint model space partition tree for the video;
- and
- generate the stabilized visual content based on the modified placement of the viewing window, the stabilized visual content including a punchout of the extents of the visual content defined by the viewing window, wherein inclusion of the extents of the visual content defined by the viewing window within the stabilized visual content effectuates stabilization of the visual content.
A system for stabilizing videos, the system comprising:
- one or more physical processors configured by machine-readable instructions to:
- obtain video information defining a video, the video including visual content having a progress length, the visual content having a source field of view;
- obtain viewing window placement information for the video, the viewing window placement information defining placement of a viewing window for the visual content as the function of progress through the progress length, the viewing window defining extents of the visual content to be included within stabilized visual content as the function of progress through the progress length, the viewing window having a punchout field of view;
- obtain constraint model space partition tree information for the video, the constraint model space partition tree information defining a constraint model space partition tree for the video, the constraint model space partition tree including branches that split a constraint model for the video, the constraint model for the video including a representation of feasible viewing window placement in which entirety of the punchout field of view of the viewing window is contained within the source field of view of the visual content;
- responsive to the placement of the viewing window outside the source field of view of the visual content, determine modified placement of the viewing window based on the constraint model space partition tree for the video;
- and
- generate the stabilized visual content based on the modified placement of the viewing window, the stabilized visual content including a punchout of the extents of the visual content defined by the viewing window, wherein inclusion of the extents of the visual content defined by the viewing window within the stabilized visual content effectuates stabilization of the visual content.
A method for stabilizing videos, the method performed by a computing system including one or more processors, the method comprising:
- obtaining, by the computing system, video information defining a video, the video including visual content having a progress length, the visual content having a source field of view;
- obtaining, by the computing system, viewing window placement information for the video, the viewing window placement information defining placement of a viewing window for the visual content as the function of progress through the progress length, the viewing window defining extents of the visual content to be included within stabilized visual content as the function of progress through the progress length, the viewing window having a punchout field of view;
- obtaining, by the computing system, constraint model space partition tree information for the video, the constraint model space partition tree information defining a constraint model space partition tree for the video, the constraint model space partition tree including branches that split a constraint model for the video, the constraint model for the video including a representation of feasible viewing window placement in which entirety of the punchout field of view of the viewing window is contained within the source field of view of the visual content;
- responsive to the placement of the viewing window outside the source field of view of the visual content, determining, by the computing system, modified placement of the viewing window based on the constraint model space partition tree for the video;
- and
- generating, by the computing system, the stabilized visual content based on the modified placement of the viewing window, the stabilized visual content including a punchout of the extents of the visual content defined by the viewing window, wherein inclusion of the extents of the visual content defined by the viewing window within the stabilized visual content effectuates stabilization of the visual content.
Dependencies are inferred from the claim text.