Canadian researchers use Bayesian algorithm for enhanced IR face recognition
SPIE reports that an algorithm based on the generalized Gaussian-mixture model offers robustness to noise and enables modeling of varying facial appearances.
According to a recent report from SPIE, infrared-based (IR) algorithms offer an alternative for handling variations in face appearance caused by illumination changes, facial expressions, and face poses during facial recognition. Although illumination and facial expression change the visual appearance of a face, thermal characteristics remain nearly unchanged. Several approaches have been proposed to analyze and recognize faces at IR wavelengths, divided into appearance- and feature-based methods. While the former focus on a face's global properties, the latter explore the facial features' statistical and geometric properties.
Researchers Tarek Elguebaly and Niza Bouguila from the Concordia Institute for Information Systems Engineering (Concordia University; Montreal, QC, Canada) have proposed an appearance-based approach using unsupervised Bayesian learning of finite, generalized Gaussian-mixture models. The scientists chose the generalized Gaussian distribution for its flexibility and because IR image statistics are generally non-Gaussian. In contrast with the expectation-maximization algorithm, Bayesian approaches consider that the data may suggest many "good" models and consider the average result computed over several models. Elguebaly and Bouguila are developing a variational framework for the learning of the generalized Gaussian-mixture model and its application to other challenges such as fusion of visual and thermal spectral modalities.
For more details and full references, visit SPIE.
-- Posted by Carrie Meadows, Vision Systems Design