Face Recognition Using Contourlet Transform and Multidirectional Illumination from a Computer Screen - Advanced Concepts for Intelligent Vision Systems - page 334

2 Subspace Feature Representation
Fig. 1 shows our image acquisition setup. For a good signal to noise ratio, the
subject must not be far from the screen. The camera's output is displayed on
the screen so that the subject can approximately center his/her face. Image
capture is automatically initiated [24] when the face is correctly positioned, or
it can be manually initiated. A white horizontal stripe scans from the top to
bottom of the screen followed by a white vertical stripe which scans from left
to right. In our experiments, the stripe was 200 pixels thick and 8 images were
captured during vertical scan and 15 during horizontal scan (given the aspect
ratio of the screen). A final image was captured in ambient light for subtracting
from all other images if required. All images are normalized so that a straight
horizontal line passes through the center of their eyes. The scale of the images is
also normalized based on the manually identified centers of eyes and lips. This
normalization is similar to the normalization used for Yale B database in [9]. The
manual identification of eyes and lips can be replaced with automatic eyes and
lips detection which can be accurately performed on the basis of all 23 images
given that they are captured instantly without subject movement. See Fig. 2 for
sample images. A mask was used to remove the lower corners of the image. We
imaged 106 subjects over a period of eight months. Out of these, 83 were imaged
in two different sessions with an average of 60 days gap.
We construct the subspaces in the feature space and use the Contourlet trans-
form [8] for extracting features. The Contourlet transform is an extension of
Wavelets. Gabor wavelets have been well studied for face recognition and many
variants exist [25][27][15]. A survey of wavelets based face recognition is given in
Fig. 1. Multiple images of a subject are acquired while illumination is varied by moving
a white stripe on a computer screen
Fig. 2. Sample faces after preprocessing