A Face Recognition under Varying Illumination - Advances in Image and Graphics Technologies - page 126

Fig. 4. Diagram of proposed method

4 Experiment and Analysis

To verify our method, we have designed an experiment with CMU-PIE. The CMU-PIE
contains a plenty of face images in many complex conditions, such as pose, illumina-
tion and expression. Since we only consider the effect of illumination to face recogni-
tion, we chose 100 front faces of 10 individuals in 10 different illumination angles from
the face database. We compared our method and the adaptive smoothing retinex. For
parameters in the experiment, we chose 40 for the iteration t and 2 layers for wavelet
transformation. All processed images are resized to100 × 100

.

Fig. 5. Examples of two methods

Fig. 5 illustrates the processed face in two methods. The first line is the original
images. The second line and third line are respectively processed by the adaptive
smoothing retinex and our method. We realized that the original images are affected by