Face recognition based on local fisher features

Abstract

To efficiently solve human face image recognition problem with an image database, many techniques have been proposed. A key step in these techniques is the extraction of features for indexing in the database and afterwards for fulfilling recognition tasks. Linear Discriminate Analysis(LDA) is a statistic method for classification. LDA filter is global in space and local in frequency. It squeezes all discriminant information into few basis vectors so that the interpretation of the extracted features becomes difficult. In this paper, we propose a new idea to enhance the performance of the LDA method for image recognition. We extract localized information of the human face images by virtue of wavelet transform. The simulation results suggest good classification ability of our proposed system.