Share Email Print
cover

Proceedings Paper

Error analysis for creating 3D face templates based on cylindrical quad-tree structure
Author(s): Weronika Gutfeter
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

Development of new biometric algorithms is parallel to advances in technology of sensing devices. Some of the limitations of the current face recognition systems may be eliminated by integrating 3D sensors into these systems. Depth sensing devices can capture a spatial structure of the face in addition to the texture and color. This kind of data is yet usually very voluminous and requires large amount of computer resources for being processed (face scans obtained with typical depth cameras contain more than 150 000 points per face). That is why defining efficient data structures for processing spatial images is crucial for further development of 3D face recognition methods. The concept described in this work fulfills the aforementioned demands. Modification of the quad-tree structure was chosen because it can be easily transformed into less dimensional data structures and maintains spatial relations between data points. We are able to interpret data stored in the tree as a pyramid of features which allow us to analyze face images using coarse-to-fine strategy, often exploited in biometric recognition systems.

Paper Details

Date Published: 11 September 2015
PDF: 10 pages
Proc. SPIE 9662, Photonics Applications in Astronomy, Communications, Industry, and High-Energy Physics Experiments 2015, 966227 (11 September 2015); doi: 10.1117/12.2205555
Show Author Affiliations
Weronika Gutfeter, Warsaw Univ. of Technology (Poland)
Research and Academic Computer Network (NASK) (Poland)


Published in SPIE Proceedings Vol. 9662:
Photonics Applications in Astronomy, Communications, Industry, and High-Energy Physics Experiments 2015
Ryszard S. Romaniuk, Editor(s)

© SPIE. Terms of Use
Back to Top