The Face Recognition Algorithm Based on Wavelet Stretching Transformation Pre-processing

A face recognition algorithm, based on wavelet stretching transformation pre-processing, has been applied to principle component analysis (PCA). PCA is one of the most important methods for feature extraction and feature recognition. The high recognition rate is based on the hypothesis that the relationship between different dimensions is linear. The gray scale of images is vulnerable to the effect of illumination which leads to an inhomogeneous gray level distribution that will decrease the image recognition rate. Wavelet-based stretching transformation pre-processing can compensate the unbalance distribution of images gray. Using wavelet transformation can decompose the image into two parts: high frequency and low frequency. In this paper both of these two parts are transformed with different functions. In order to illustrate the effect of the pre-processing algorithm, two comparison experiments are presented to show its validity. The training and testing samples are selected with random sequence from ORL and Feret database. The experimental results show that after pre-processing with stretching transformation, the achieved face recognition rate is 96% in ORL database and 93% in Feret database. Therefore, this fact demonstrates that wavelet-based stretching transformation has advantage of achieving higher accuracy of recognition than the traditional ones.