Integrated 3D Scene Flow and Structure Recovery from Multiview Image Sequences

Scene ow is the 3D motion eld of points in the world. Given N (N > 1) image sequences gather ed with a N-eye stereo camera or N calibrated cameras, we present a novel system which integrates3D scene ow and structur erecovery in order to complement each other's performance. We do not assume rigidity of the sc ene motion, thus allowing for non-rigid motion in the scene. In our work, images are segmented into small regions. We assume that each small region is undergoing similar motion, represente dby a 3D a ne model. Non-linear motion model tting basedon both optical ow constr aintsand stereo constr aintsis then carriedover each image region in order to simultaneously estimate 3D motion correspondences and structure. T oensur ethe robustness, several regularization constr aints are also introduced. A recursive algorithm is designed to incorporatethe local and regularization constr aints.Experimental results on both synthetic and real data demonstrate the e ectiveness of our integ...

• Type: Conference
• Year: 2000
• Where: CVPR
• Authors: Ye Zhang, Chandra Kambhamettu
• Keywords: 3D Motion Eld, Computer Vision, CVPR 2000, Integrated 3D Motion, Integrates3D Scene Ow, Non-linear Motion Model, Sc Ene Motion