Seminar: Low-cost Realistic 3D Face Reconstruction

Zhang Chen

LSU Visiting Scholar, Shanghai Tech University, School of Information Science and Technology

September 28, 2018

3:00 pm

Patrick F. Taylor Hall 3107

Abstract

The digitization of photorealistic 3D face is a long-standing problem in computer vision and graphics. Successful solutions can benefit numerous applications, ranging from visual effects production to face recognition. Human faces contain both coarse structures (e.g., nose, cheek, lip, forehead) and fine details (e.g., wrinkles, eyebrows, beards, and pores). Recovering 3D face models with high-frequency details is a challenging task that generally requires expensive equipment and high computational effort.

In this talk, I’ll present a high-precision but low-cost solution for realistic 3D face reconstruction. We first recover a coarse 3D face from a single image using morphable face models and then refine the fine details through an iterative optimization. Our technique is able to deliver high quality 3D face reconstruction with ultra-fine details such as wrinkles, eyebrows, whelks, pores, etc. using a cheap and easy-to-built hardware setup. The reconstruction quality is comparable to movie quality productions captured by expensive instruments.

Bio

Zhang Chen is a Ph.D. student in School of Information Science and Technology at ShanghaiTech University, China. He is currently a visiting scholar at Louisiana State University with Dr. Jinwei Ye. His research interests lie in computer vision and graphics with focus on 3D reconstruction techniques. Zhang received his bachelor degree in Electrical and Computer Engineering from Shanghai Jiao Tong University, China in 2014.