Comparison between Thermal and Visible Facial Features on a Verification Approach - Biomedical Engineering Systems and Technologies - page 276

heat-emitting objects are present. In this approach, the distance from centroid (DFC)
shows its suitability for comparing the degree of symmetry of the lower face outline.
The use of correlation filters in [6] has shown its adequacy for face recognition
tasks using thermal infrared (IR) face images due to the invariance of this type of
images to visible illumination variations. The results with Minimum Average Correla-
tion Energy (MACE) filters and Optimum Trade-off Synthetic Discriminant Function
(OTSDF) in low resolution images (20x20 pixels) prove their efficiency in Human
Identification at a Distance (HID).
Scale Invariant Feature Transform (SIFT) algorithm [7] are widely used in object
recognition. In [8], SIFT has appeared as a suitable method to enhance the recognition
of facial expressions under varying poses over 2D images It has been demonstrated
how affine transformation consistency between two faces can be used to discard SIFT
mismatches.
Gender recognition is another lively research field working with SIFT algorithm.
In [9], faces are represented in terms of dense-Scale Invariant Feature Transform (d-
SIFT) and shape. Instead of extracting descriptors around interest points only, local
feature descriptors are extracted at regular image grid points, allowing dense descrip-
tions of face images.
However, systems generate large number of SIFT features from an image. This
huge computational effort associated with feature matching limits its application to
face recognition. An approach to this problem has been developed in [10], using a
discriminating method. Computational complexity is reduced more than 4 times and
accuracy is increased in 1.00% on average by checking irrelevant features.
Constructing methods that scale well with the size of a database and allow finding
one element of a large number of objects in acceptable time is an avoidable challenge.
This work is inspired by Nister and Stewenius [11], where object recognition by a k-
means vocabulary tree is presented. Efficiency is proved by a live demonstration that
recognized CD-covers from a database of 40000 images. The vocabulary tree showed
good results when a large number of distinctive descriptors form a large vocabulary.
Many different approaches to this solution have been developed in the last few years
[12] and [13], showing its competency organizing several objects. Having regard to
these good results, this solution will be tested in this paper with SIFT descriptors in a
vocabulary tree.
In this context, the aim of this work is to propose, compare and evaluate a facial
and head verification system for visible and thermal ranges, applying the SIFT algo-
rithm and obtaining local distinctive descriptors from each image based on [14]. The
construction of the vocabulary tree enables to have these descriptors hierarchically
organized and ready to carry out a search to find a specific object.
This paper is organized as follows. The proposed system is presented in section 2.
Description of experiments and results are outlined in section 4. Discussions are de-
scribed in section 4. Finally, conclusions are given in section 5.

2 Facial Recognition Approach

In our system proposed, SIFT descriptors are used to extract information from thermal
and visible images in order to verify the identity of a test subject. Local distinctive
descriptors are obtained from each face in the database and are used to build a