New Algorithm Compares 3D Faces Faster Using "Average Face"

New algorithm uses average face to compare 3D faces fast.

Why it matters

  • New algorithm compares 3D faces much faster than before.
  • Uses "average face" to speed up the process.
  • Handles holes and edges in scans, making it robust.

By the numbers

  • Datasets: 100, 500, and 1000 faces.
  • Speedup: hours to minutes, thousands of times faster.

The big picture

  • Uses average face to align and compare other faces.
  • Addresses real-world 3D scan issues like holes and edges.

What they're saying

  • Technique similar to "eigenfaces" and PCA.
  • Concerns about surveillance technology misuse.
  • Praise for handling noisy facial data effectively.

Caveats

  • Tested on facial scans; may not suit other 3D models.
  • Depends on polygonal meshes; may struggle with occlusions or head poses.

What’s next

  • Automate template face selection.
  • Improve pre-alignment strategies.
  • Address occlusion and head pose issues.