In [15], Huang et al. proposed to scan wrinkled 3D face models
to create blendshape models, and successfully reconstruct the
Example-based texture synthesis: It aims to synthesize a new
texture appearing to be generated by the same process of the given
exemplar. The 2D texture synthesis algorithms can be categorized
into pixel-level algorithms [17,18], i.e. synthesize textures at each
pixel, and patch-level algorithms, which can synthesize textures
at a connected group of pixels simultaneously [19,20]. They are
generally faster than pixel-based approaches.