AUTOMATIC HUMAN AGE ESTIMATION BASED ON NEURAL NETWORKS AND THE MODIFIED FACE MODEL

摘要

Ideally image features for recognition systems such as age estimation should only depend on the structure of the facial images. So in task of age estimation, we introduce some changes in the previous face model able to compensate lighting condition of face images. In this face model, the intensity model has been replaced by the structure model that applies the features generated by the modified census transform than gray-level intensities. Based on this face model, improvement of the results compared to the previous face model is achieved. The effectiveness of our face model has been successfully tested on age estimation using 1002 FG-NET aging face images corresponding to 82 subjects, which were acquired under variable illumination conditions.