Hyperlapse video through time lapse and stabilization

Abstract

Various techniques described herein involve generating output hyper-lapse video from input video. A smoothed camera path can be calculated based on the input video. Additionally, an output camera pose may be selected from the smoothed camera path for outputting an output frame of hyper-lapse video. One or more selected input frames may be picked from the input video for output frames. The selected input frame may be chosen based at least in part on the output camera pose for the output frame. Additionally, selected input frames can be combined to render an output frame. Picking selected input frames from the input video and combining the selected input frames may be performed for each of the output frames of the output hyper-lapse video.

Description

Background technique [0001] Conventional cameras are often expensive, bulky and difficult to operate devices. Recently, digital cameras have become cheaper, smaller, easier to use and have become virtually ubiquitous. Cameras are now often attached to cars, computers, and phones. In addition, wearable cameras are becoming more common. A wearable camera may be referred to as a first-person camera because the action is seen through the eyes of the camera operator. [0002] The captured video can be long and monotonous, and thus, can be tedious to watch and difficult to navigate. Various conventional methods employ time-lapse techniques to shorten such videos. For example, the speed of a video can be increased by selecting every nth frame of the video for the resulting time-lapse video. [0003] Additionally, the camera can be used to capture video while in motion. Traditional time-lapse techniques can be applied to video captured by a camera moving in space. The resulting...

AI Technical Summary

Problems solved by technology

Thus, conventional time-lapse techniques that increase the frame rate of such videos can amplify camera shake, thereby reducing the viewability of the resulting video

Images