A New Approach to Face Image Coding using Gabor Wavelet Networks

Abstract

In this paper, we describe a new approach to face image coding using Gabor Wavelet Networks (GWN). This type of network yields good results in many signal coding applications and was already used for face representation by Kruger in 2001. The main idea is to approximate the face image, considered as a two dimensional function, with a set of Gabor wavelets. We describe an improved network training technique, which uses several face images of the same person rather than a single one. This technique yields the model of the face image, but also statistical criteria defining the significance of each Gabor wavelet in this model.