eduzhai > Applied Sciences > Engineering >

Face Super-Resolution Guided by 3D Facial Priors

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 16 pages

Abstract: State-of-the-art face super-resolution methods employ deep convolutionalneural networks to learn a mapping between low- and high- resolution facialpatterns by exploring local appearance knowledge. However, most of thesemethods do not well exploit facial structures and identity information, andstruggle to deal with facial images that exhibit large pose variations. In thispaper, we propose a novel face super-resolution method that explicitlyincorporates 3D facial priors which grasp the sharp facial structures. Our workis the first to explore 3D morphable knowledge based on the fusion ofparametric descriptions of face attributes (e.g., identity, facial expression,texture, illumination, and face pose). Furthermore, the priors can easily beincorporated into any network and are extremely efficient in improving theperformance and accelerating the convergence speed. Firstly, a 3D facerendering branch is set up to obtain 3D priors of salient facial structures andidentity knowledge. Secondly, the Spatial Attention Module is used to betterexploit this hierarchical information (i.e., intensity similarity, 3D facialstructure, and identity content) for the super-resolution problem. Extensiveexperiments demonstrate that the proposed 3D priors achieve superior facesuper-resolution results over the state-of-the-arts.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×