Identification of the Background from Video Based on Motion Constraints
YU Yong-yan, Yuqin Shu · Advances in computer science research · 2015
To identity of the background from video is one of the most important topic in 3D reconstruction about dynamic scenes,which would give a lot of information and give the simplest interpretation for the overall scene motion.This paper attempt to lift that limitation and the proposed techniques are based on the fact that the aforementioned motion characteristics of an object are lost not only in the case of wrong relative scales but also in the case of a wrong background identification,so will propose two techniques based on the independence criterion and the non-accidentalness principle.The independence criterion is applicable to any scene, however requires many frames to be statistically valid and needs variation in the motion parameters.