Coupled reconstruction of hair and skin

摘要

Techniques are provided to model hair and skin. Multiscopic images are received that depict at least part of a subject having hair. The multiscopic images are analyzed to determine hairs depicted. Two-dimensional hair segments are generated that represent the hairs. Three-dimensional hair segments are generated based on the two-dimensional hair segments. A three-dimensional model of skin is generated based on the three-dimensional hair segments.

主权项

1. A computer-implemented method to model facial hair and skin, the computer-implemented method comprising:
receiving a plurality of multiscopic images depicting at least part of a face of a subject having facial hair; analyzing the plurality of multiscopic images in order to determine correspondences between the plurality of multiscopic images and further determine one or more facial hairs depicted in the plurality of multiscopic images; generating a disparity map based on the correspondences, the disparity map containing one or more disparity values adjusted in directions of improved photometric-consistency and improved surface-consistency, respectively; generating a geometry mesh of the subject, based on the disparity map; generating one or more 3D facial hair segments representing the determined one or more facial hairs; generating a 3D model of facial skin underlying the one or more facial hairs, based on the generated one or more 3D facial hair segments and by operation of one or more computer processors; and generating a 3D model of the subject, based on the geometry mesh, the one or more 3D facial hair segments, and the 3D model of facial skin.