Stabilizing video using transformation matrices

One or more non-transitory computer-readable devices including instructions that, when executed by one or more processors, cause performance of operations that include:

• receiving, by a computing system, first and second frames of a video that was captured by a camera;
• generating, by the computing system and using the first and second frames of the video, a first mathematical transformation that indicates movement of the camera with respect to a scene captured by the video from when the first frame was captured to when the second frame was captured, including movement that began recently and movement that has been occurring over a longer term, the computing system generating the first mathematical transformation by identifying how features points that are present in the first frame and the second frame moved between the first frame and the second frame;
• identifying, by the computing system, movement of the camera with respect to the scene that has been occurring over the longer term;
• generating, by the computing system using the first mathematical transformation and the identified movement of the camera with respect to the scene that has been occurring over the longer term, a second mathematical transformation by a process that involves removing from the first mathematical transformation indications of the identified movement of the camera that has been occurring over the longer term, the second mathematical transformation being more representative of movement of the camera with respect to the scene that began recently and less representative of movement of the camera with respect to the scene that has been occurring over the longer term, in comparison to the first mathematical transformation;
• generating, by the computing system, a stabilized version of the second frame by applying the second mathematical transformation to the second frame to remove from the second frame movement that began recently and leave movement that has been occurring over the longer term, without taking into account movement of the camera with respect to the scene from future frames of the video.