Capture of dynamic scene using multiple cameras provides rich spatial-temporal information

Опубликовано 15 августа 2016, 18:28

"Capture of dynamic scene using multiple cameras provides rich spatial-temporal information that can be used for solving challenging computer vision problems.
Indeed, multi-view systems have been used by various computer-vision methods in the last decade, where the setting of cameras depends on the application. In this talk, I will discuss the use of multi-view systems in two “distant” setups: First, I will present the use of calibrated and synchronized camera-array for solving the problem of dense 3D structure and 3D motion estimation. Then, I will consider the scenario of a group of people capturing asynchronously a dynamic scene, using probably the most popular means of photographing today – cellphones. The combined data obtained this way can be regarded as the output of a new type of extended camera, which we call a crowd-based camera (or CrowdCam). Using this new setup introduced us with a novel problem of photo-sequencing -- temporally ordering a set of still images taken by a CrowdCam, and a geometry-based solution, as I will present. We believe that photo sequencing is an essential step for analysis of a dynamic scene form a set of still images."