Research On Multi-directional LBP Based Face Recognition Algorithm And Its Realization

More and more attentions have been paid to face recognition because of its wide application scope. However, face recognition is a challenging task since the speciality of human face pattern and diversity and complexity of impact from external factors, such as illumination.This paper dedicated to the study of classical PCA algorithm and some other excellent feature extraction algorithm, such as LBP. PCA algorithm can be used to extract global feature which contains information of pattern relationship between samples, but it is weak in extract local feature of human faces. LBP depicts facefeatures in three levers------pixel lever, subregion lever, image lever. LBP features isrobust to illumination and rotation, additionally, it is easy to calculate. LBP is used for extract local feature of image, such as edge, texture which with directional information. However, LBP algorithm can not extract enough directional information. Regarding to this, this paper proposed a multi- directional LBP algorithm, which is more advanced in extracting directional information. Experiment were conducted on ORL face database and the result demonstrate the effective of our algorithm.Base on this algorithm cooperate with PCA and ridge regression, we constructed an real-time face recognition system using VC++6.0 and OpenCV. This system can detect and recognize human faces from the input video frame real-timely. This system provide an reference for us to developing a more practical face recognition system in the future research.