Statistical multiple light source detection

Christos-Savvas Bouganis, Mike BrookesIET 2007 · DOI

Abstract

Multiple light source detection has many applications in image synthesis and augmented reality. Current techniques can provide accurate results but have limited applicability in real-life scenarios where interaction with the scene is not possible. The authors provide a statistical framework for multiple light source detection that relies on the common features of objects belonging to a particular class and illustrate it using the class of human faces. Experiments with real data demonstrate that a light distribution with up to three light sources can be detected within 13° mean error. Application of the proposed framework to the problem of 3D reconstruction from multiple images under arbitrary lighting demonstrates the effectiveness of the framework compared with current techniques.
BibTeX
@inproceedings{bouganis2007statistical,
  author = {Christos-Savvas Bouganis and Mike Brookes},
  title = {{Statistical multiple light source detection}},
  booktitle = {IET},
  year = {2007},
  doi = {10.1049/iet-cvi:20065001},
}

Loading related content...

Figure