A Parametric Model for Human Faces

How to read this

Category

PhD dissertation, the founding parametric face model

Contributions

• Introduces a parametric 3D face model driven by a small set of continuous conformation parameters, controlling identity or shape such as face width and jaw shape, and separate expression parameters, controlling things like eyelid opening or smile
• Replaces pose interpolation between digitized face scans with direct parametric control of a polygonal mesh, letting a small number of sliders generate a continuous range of expressions rather than blending between a fixed set of captured poses
• Establishes the conformation versus expression parameter split that essentially every later rig based facial animation approach, blendshapes and FACS driven rigs included, still conceptually inherits

Context

This dissertation extends Parke's own 1972 work, Computer Generated Animation of Faces, which animated faces by interpolating between digitized poses of a real face. It is the acknowledged root of the digital face lineage that runs through Parke's own 1982 IEEE Computer Graphics and Applications paper, through Ekman and Friesen's FACS as a shared vocabulary layer, and on into the blendshape and muscle based facial rigs used throughout film and games today.Builds on: Computer Generated Animation of Faces

Correctness

The copy of this dissertation in the archive is a scanned image with no extractable text layer, so this guide is grounded in the well documented history of the work and Parke's own later published summaries of it rather than a direct read of this specific scan. As a piece of research it was a 1970s proof of concept, a handful of parameters driving a low polygon mesh on period hardware, not a validated perceptual study or production system, and it necessarily looks primitive by any current standard.

Clarity

Written as a 1970s computer science dissertation, dense and dated in style, more valuable to know about than to read start to finish.

How to read it

Read only the abstract and conclusion for the parametric idea itself, that is the historically important part. For the actual mechanics, Parke's own 1982 paper, Parameterized Models for Facial Animation, restates the same ideas more concisely and is a better use of reading time. Treat this 1974 dissertation as a citation worth knowing, not a working reference to implement a rig from.