Face recognition using a fourier polar based approach

Abstract

In this work, a face recognition approach for human identification is proposed. It is based on the Discrete Fourier Transform of the original spatial image and on the polar mapping of the spectrum of the latter image. While the first step eliminates the translation's effect between faces, the second step minimizes the effect of any face's rotation. The approach is evaluated by studying its performance with other existing techniques using a correlation coefficient as a similarity measure for comparison purposes. Besides, the effect of the number of images used in the training procedure is studied. The results show that the proposed technique outperforms the other techniques and a recognition of 100 percent is possible as the number of training images is increased.