An effective preprocessing scheme for face recognition based on local Gabor binary pattern histogram sequence

To a great extent, the performance and robustness of automated face recognition systems are impacted by within-class variations between gallery and probe images acquired in a variety of conditions. In particular, changes of illumination and facial expression contribute mainly to these intrapersonal variations of facial images. In this paper, an effective preprocessing scheme is proposed for face recognition based on local Gabor binary pattern histogram sequence (LGBPHS). The scheme to normalize within-class variations incorporates a Gamma correction transformation and a local normalization procedure. The experimental results obtained on FERET face database show that the preprocessing scheme significantly improves the recognition performance, superior to several existing preprocessors. Furthermore, in the comparisons our method impressively outperforms state-of-art face identification techniques.