A face recognition method robust to partial illumination

In this work, a face recognition method that is robust to partial illumination is introduced. The proposed method applies the devised preprocessing algorithm in order to avoid errors raised by partial illumination when it decides it is necessary. The rapidness of this preprocessing algorithm and applying it only when it is necessary reduce its effect on the speed of recognition. The proposed face recognition approach exploits the features found by Haar Discrete Wavelet Transform. These features are stored to a database and recognition is made by comparing the features of the sample under the test with those of recorded samples according to the L1 distance. Storing every sample of each subject and their various illumination models applied versions increased the recognition rate in a considerable manner. The tests made by using the ORL and the Yale Face databases give 96% and 94% successes respectively.