Clothed and naked human shapes estimation from a single image

Abstract

This paper presents a novel method to simultaneously estimate the clothed and naked 3D shapes of a person. The method needs only a single photograph of a person wearing clothing. Firstly, we learn a deformable model of human clothed body shapes from a database. Then, given an input image, the deformable model is initialized with a few user-specified 2D joints and contours of the person. And the correspondence between 3D shape and 2D contours is established automatically. Finally, we optimize the parameters of the deformable model in an iterative way, and then obtain the clothed and naked 3D shapes of the person simultaneously. The experimental results on real images demonstrate the effectiveness of our method. © 2012 Springer-Verlag.