Simulated video with extra viewpoints and enhanced resolution for traffic cameras
AssigneeMicrosoft Corporation
InventorsBilly Chen, Eyal Ofek
Life of this patent
Estimating the term…
Abstract
Simulated high resolution, multi-view video based on video input from low resolution, single-direction cameras is provided. Video received from traffic cameras, security cameras, monitoring cameras, and comparable ones is fused with patches from a database of pre-captured images and/or temporally shifted video to create higher quality video, as well as multiple viewpoints for the same camera.
Claims20
A method to be executed at least in part in a computing device for generating an enhanced simulated video from a camera feed, the method comprising:
- receiving an original video input from a single-direction camera;
- generating a stream of vehicle video patches;
- matching vehicles in the vehicle video patches to corresponding stored images of vehicles;
- retrieving images of a different angle corresponding to the stored images of the vehicles;
- fusing the retrieved images into a simulated video displaying the vehicles moving in another direction;
- and
- rendering the simulated video.
A system for generating an enhanced simulated video from a camera feed, the system comprising:
- at least one single-directional camera;
- a server coupled to the at least one single-directional camera, the server executing a video processing application, wherein the video processing application is configured to:
- receive an original video input from the single-direction camera;
- generate a stream of vehicle video patches;
- buffer a predefined number of video frames;
- capture images of a first set of vehicles in frames where the vehicles appear near the camera;
- superimpose the images over frames where the first group of vehicles appear away from the camera;
- match a second group of vehicles in the vehicle video patches to corresponding stored images of vehicles;
- retrieve images of a different angle corresponding to the stored images of the vehicles;
- fuse the superimposed images into a first simulated video with enhanced resolution;
- fuse the retrieved images into a simulated second video displaying the second group of vehicles moving in another direction;
- and
- render the simulated videos.
A computer-readable storage medium with instructions stored thereon for generating an enhanced simulated video from a camera feed, the instructions comprising:
- receiving an original video input from a single-direction camera;
- generating a stream of vehicle video patches;
- buffering a predefined number of frames;
- matching the buffered frames employing normalized cross-correlation such that images of vehicles farther away from the camera are enhanced;
- and
- fusing the matched frames into the simulated video.
Dependencies are inferred from the claim text.