Face Recognition in the Thermal Infrared Spectrum

Abstract

We present a two-stage face recognition method based on infrared imaging and statistical modeling. In the first stage we reduce the search space by finding highly likely candidates before arriving at a singular conclusion during the second stage. Previous work has shown that Bessel forms model accurately the marginal densities of filtered components and can be used to find likely matches but not a unique solution [1]. We present an enhancement to this approach by applying Bessel modeling on the facial region only rather than the entire image and by pipelining a classification algorithm to produce a unique solution. The detailed steps of our method are as follows: First, the faces are separated from the background using adaptive fuzzy connectedness segmentation. Second, Gabor filtering is used as a spectral analysis tool. Third, the derivative filtered images are modeled using two-parameter Bessel forms. Fourth, high probability subjects are short-listed by applying the L 2 -norm on the Bessel models. Finally, the resulting set of highly likely matches is fed to a Bayesian classifier to find the exact match. We show experimentally that segmentation of the facial regions results in better hypothesis pruning and classification performance. We also present comparative experimental results with an eigenface approach to highlight the potential of our method.

Structure from Infrared Stereo Images

“…Other researchers, including [79], [11] have used feature-based approaches (rather than appearance-based methods), to overcome the challenging conditions such as variable poses and facial expressions. Buddharaju et al in [11] used spectral features.…”

“…Buddharaju et al in [11] used spectral features. Their method prunes the hypothesis space by modeling the extracted spectral features through Bessel parametric forms.…”

Adaptive Pointing Theory (APT) Artificial Neural Network

“…Then largest component has been extracted from a binary image using \Connected Component Labeling" algorithm. 9 This algorithm is based either on \4-connected" neighbors or \8-connected" neighbor's method. 7,17 In this study \8-connected" neighbor's method is used.…”

Blood Perfusion Models for Infrared Face Recognition

“…Furthermore, as discussed by Prokoski (Prokoski, 2000), a facial thermal pattern is determined by the vascular structure of each face, which is irreproducible and unique. Based on the assumption that facial thermal patterns are determined by blood vessels transporting warm blood, Prokoski tried to extract the blood vessel minutiae (Prokoski, 2001) or vascular network (Buddharaju et al, 2004, Buddharaju et al, 2005 as the facial features for recognition. The basic idea is to extract such features using image segmentation.…”

“…It has been indicated by Guyton & Hall (Guyton & Hall, 1996) that the average diameter of blood vessels is around 10~15μm, which is too small to be detected by current IR cameras (limited by the spatial resolution); the skin directly above a blood vessel is on average 0.1°C warmer than the adjacent skin, which is beyond the thermal accuracy of current IR cameras. The methods using image segmentation in (Prokoski, 2001, Buddharaju et al, 2004, Buddharaju et al, 2005 are heuristic, and it still remains a big challenge to capture the pattern of blood vessels on each face. On the other hand, the phenomenon of " homoiotherm" due to human temperature regulation has led to the direct use of thermograms for recognition (Wilder et al, 1996, Socolinsky & Selinger, 2002, Wu et al, 2003, Chen et al, 2005.…”