Microsoft Research Asia advances facial recognition software
10 August, 2011
Researchers from Microsoft Research Asia are presenting new methods of capturing high-fidelity 3D facial expressions at SIGGRAPH 2011 in Vancouver.
The research team, made up of Haoda Huang, Xin Tong and Hsiang-Tao Wu from Microsoft Research Asia, along with Jinziang Chai from Texas A&M University, used a two-step process to create more realistic 3D images.
AsianScientist reports that the team used both marker-based motion capture and high-resolution scanners to recreate faces that include details such as wrinkles that move as the image moves. Using the dual process meant that a minimal number of scans were necessary for accurate facial reconstruction.
While the team still needs to tweak its approach in order to fully realize its goals, this research may prove to be beneficial in the fields computerized film making and avatar creation, in terms of making more realistic images.
Read more here.