Abstract:

A novel algorithm based on two-dimensional discrete wavelet transform for frontal face detection was presented. The algorithm calculated waveletfaces by using Haar wavelet, deduced corresponding formulation for face feature extraction, and used perceptron algorithm to train linear classifier for classification. The comparison between the proposed approach and Eigenfaces method was carried out on four different face data sets. The results show that the computational efficiency and detection accuracy of the proposed algorithm is superior to Eigenfaces method.

Key words: Face detection, discrete wavelet transform, waveletfaces