Bilinear Kernel Reduced Rank Regression for Facial Expression Synthesis

In the last few years, Facial Expression Synthesis (FES) has been a flourishing area of research driven by applications in character animation, computer games, and human computer interaction. This paper proposes a photorealistic FES method based on Bilinear Kernel Reduced Rank Regression (BKRRR). BKRRR learns a high-dimensional mapping between the appearance of a neutral face and a variety of expressions (e.g. smile, surprise, squint). There are two main contributions in this paper: (1) Propose BKRRR for FES. Several algorithms for learning the parameters of BKRRR are evaluated. (2) Propose a new method to preserve subtle person-specific facial characteristics (e.g. wrinkles, pimples). Experimental results on the CMU Multi-PIE database and pictures taken with a regular camera show the effectiveness of our approach.

• Type: Conference
• Year: 2010
• Where: ECCV
• Keywords: Computer Vision, ECCV 2010, Facial Expression Synthesis, Kernel Reduced Rank, Photorealistic Fes Method