Facial Performance Synthesis using Deformation-Driven Polynomial Displacement Maps

How to read this

Category

Method: data-driven facial detail synthesis

Contributions

• Polynomial displacement maps (PDMs) that encode the mapping from sparse motion-capture markers to high-resolution face geometry
• A real-time 3D scanning setup that captures facial performance down to wrinkle and pore detail
• Synthesis of novel performances with realistic medium- and fine-scale skin detail from coarse marker input

Context

Builds on high-resolution facial capture in the lineage of Acquiring the Reflectance Field of a Human Face (Debevec et al.), turning captured detail into a learned marker-to-geometry deformation model.Builds on: Acquiring the Reflectance Field of a Human Face

Correctness

Demonstrated on multiple subjects and expressions, generating detailed geometry from coarse markers; readers should note the polynomial map is fit per-subject from captured data, so extrapolation beyond the captured expression range and transfer across subjects are limitations.

Clarity

Accessible; a first pass conveys the marker-driven detail idea, a second pass clarifies the polynomial map fitting.

How to read it

Focus on how marker motion parameterizes the displacement maps and the multi-scale (medium and fine) split; a second pass on the PDM fitting pays off if you want to reproduce wrinkle synthesis.