Real-Time Avatar Pose Transfer and Motion Generation Using Locally Encoded Laplacian Offsets

(整期优先)网络出版时间:2019-02-12
/ 1
Weproposeahumanavatarrepresentationschemebasedonintrinsiccoordinates,whichareinvarianttoisometryandinsensitivetohumanposechanges,andanefficientposetransferalgorithmthatcanutilizethisrepresentationtoreconstructahumanbodygeometryfollowingagivenpose.Suchaposetransferalgorithmcanbeusedtocontrolthemovementofanavatarmodelinvirtualrealityenvironmentsfollowingauser'smotioninrealtime.Ourproposedalgorithmconsistsofthreemainsteps.First,werecognizetheuser'sposeandselectatemplatemodelfromthedatabasewhohasasimilarpose;then,theintrinsicLaplacianoffsetsencodedinlocalcoordinatesareusedtoreconstructthehumanbodygeometryfollowingthetemplatepose;finally,themorphingbetweenthetwoposesisgeneratedusingalinearinterpolation.Weperformexperimentstoevaluatetheaccuracyandefficiencyofouralgorithm.Webelieveourproposedsystemisapromisinghumanmodelingtoolthatcanbeusedingeneralvirtualrealityapplications.