Facetransfer: a system model of facial image rendering

This paper describes a system model for facial image rendering of a prototype system - Facetransfer, which is developed to synthesize realistic expression, age, and view of face images. The Facetransfer system integrates the linear object class model, shape-texture distinctiveness model, high-level attributes clone model and view space projection model into a fusion model to analyze and render 2D human face images bypassing the re-building of 3D models. We demonstrate the attractive features of our system with some examples with different races from AI&R Asian face database and MPI caucasian face database. The Facetransfer system driven by multiple face databases can be widely applied to criminal detective, visualization, artistic rendering, and multimedia, etc.