Authors:
Jasenko Zivanov
;
Pascal Paysan
and
Thomas Vetter
Affiliation:
University of Basel, Switzerland
Keyword(s):
Face rendering, Face normal map, Normal map capture, Surface reconstruction.
Related
Ontology
Subjects/Areas/Topics:
Applications
;
Computer Vision, Visualization and Computer Graphics
;
Geometric Computing
;
Geometry and Modeling
;
Image-Based Modeling
;
Image-Based Rendering
;
Lighting and Appearance
;
Pattern Recognition
;
Reflection and Illumination Models
;
Rendering
;
Software Engineering
;
Surface Modeling
;
Texture Models, Analysis, and Synthesis
Abstract:
Obtaining photorealistic scans of human faces is both challenging and expensive. Capturing the highfrequency components of skin surface structure requires the face to be scanned at very high resolutions, outside the range of most structured light 3D scanners.
We present a novel and simple enhancement to the acquisition process, requiring only four photographic flashlights and three texture cameras attached to the structured light scanner setup.
The three texture cameras capture one texture map (luminance map) of the face as illuminated by each of the four flash-lights. Based on those four luminance textures, three normal maps of the head are approximated, one for each color channel. Those normal maps are then used to reconstruct a 3D model of the head at a much higher mesh resolution, in order to validate the normals. Finally, the validated normals are used as a normal map at rendering time. Alternatively, the reconstructed high resolution model can also be used for rendering.