Local variance projection log energy entropy features for illumination robust face recognition

This paper presents local variance projection log energy entropy (LVP-LEE) features for illumination robust face recognition. In the proposed method, one face image is firstly divided into non-overlapping small blocks, and then the vertical and horizontal variance projection distributions of each small face block are computed, using to extract features in terms of log energy entropy concept. Finally, features from individual block are constructed as the whole feature vector of face image for face recognition. The proposed LVP-LEE features have been evaluated based on the CMU PIE face database and compared with PCA, modular PCA and sub pattern PCA features. Experimental results show that the proposed LVP-LEE features are more effective for illumination robust face recognition.