Fast deformable model-based human performance capture and FVV using consumer-grade RGB-D sensors
作者：
Highlights：
• A novel end-to-end system for fast model-based reconstruction of human actor performances into 3D mesh sequences is proposed.
• The system takes as input the RGB-Depth data from multiple low-cost sensors.
• It outputs a sequence of temporally-coherent textured meshes, enabling realistic Free Viewpoint Video (FVV).
• The system implicitly performs human articulated motion tracking.
• The framework is based on the automatic generation of a skinned model, on-line pose estimation and model deformation.
摘要
•A novel end-to-end system for fast model-based reconstruction of human actor performances into 3D mesh sequences is proposed.•The system takes as input the RGB-Depth data from multiple low-cost sensors.•It outputs a sequence of temporally-coherent textured meshes, enabling realistic Free Viewpoint Video (FVV).•The system implicitly performs human articulated motion tracking.•The framework is based on the automatic generation of a skinned model, on-line pose estimation and model deformation.
论文关键词：Dynamic mesh sequences,RGB-D sensors,3D reconstruction,Articulated motion,Surface deformation,Deformable models,Free Viewpoint
论文评审过程：Received 11 December 2016, Revised 19 December 2017, Accepted 11 February 2018, Available online 13 February 2018, Version of Record 23 February 2018.
论文官网地址：https://doi.org/10.1016/j.patcog.2018.02.013
