DeReEs: real-time registration of RGBD images using image-based feature detection and robust 3D correspondence estimation and refinement

摘要

We present DeReEs, a real-time RGBD registration algorithm for the scenarioudwhere multiple RGBD images of the same scene are obtained from depth-sensingudcameras placed at different viewpoints, with partial overlaps between their views.udDeReEs (Detection, Rejection and Estimation) is a combination of 2D image-basedudfeature detection algorithms, a RANSAC based false correspondence rejection and audrigid 3D transformation estimation. DeReEs performs global registration not only inudreal-time, but also supports large transformation distances for both translations andudrotations. DeReEs is designed as part of a virtual/augmented reality solution for audremote 3D collaboration system that does not require initial setup and allows users toudfreely move the cameras during use. We present comparisons of DeReEs with otherudcommon registration algorithms. Our results suggest that DeReEs provides betterudspeed and accuracy especially in scenes with partial overlapping.