Performance evaluation of LBP and LDP rotated and scaled faces

Face detection and recognition is becoming increasingly important in context of surveillance, credit card fraud detection, assistive devices for visual impaired, etc. There are multiple cues available and can be used as features. It has been observed that local features perform better than global cues. Local binary pattern and local derivative pattern try to encode directional pattern of a face image. Hence it becomes important to critically evaluate the performance of LBP and LDP for rotated and scaled faces. Facial features are extracted and compared using support vector machine classification algorithm. We have considered standard databases containing the rotated faces. Also, we have created a database of annotated rotation angle to find out the permissible degree of rotation for LBP and LDP.