Method for computing visual field of multi-view dynamic scene acquisition

Qionghai Dai,Yebin Liu,Di Wu

2012-01-01

Abstract:The embodiment of the invention discloses a method for computing a visual field in multi-view dynamic scene acquisition of time decoupling, which comprises the following steps: constructing a system for acquiring multi-view dynamic scenes; sampling visual fields of the dynamic scenes in a view-time two-dimensional visual field sub-space combining with a spatial interleaving sampling strategy, a time interleaving exposure sampling strategy and a variable exposure sampling strategy; based on the unity and coherence of the time intersection and the space intersection in a time decoupling strategy, intersecting spatial outlines of the dynamic scenes performed with time-space interleave exposure to obtain a scene model corresponding to a high time resolution under the condition of short-time exposure; back-projecting the outline of the scene model to acquire an outline motion restraint to perform the time decoupling to remove motion blur; establishing a three-dimensional model of the dynamic scene of which the motion blur is removed by decoupling through a three-dimensional reconstruction algorithm; and performing alternating iterative optimization on the three-dimensional model of thescene and time decoupling images under each short-time exposure. The scheme provided by the invention can capture the high-speed dynamic of the three-dimensional scene in real world without an expensive industrial camera, is a universal method for sampling visual field, and has very high practical value and adaptability.