Comparison of Various Local Binary Pattern Operators

This paper proposes and presents generalizations to the gray scale and rotation invariant texture classification face representation methods LBP, ILBP, SLBP and MBLBP based on local binary patterns. The average recognition rate of face recognition system depends on the representation method used for face images. The face representation method should be robust against to the illumination variations. Both gray-level images and Gabor feature images are used to evaluate the comparative performances of various LBP operators. Extensive experimental results on FERET, Extended Yale B, databases of sample images with respect to the parameters Equal Error Rate (EER), computational time show that the SLBP is faster than other LBP methods and EER is less using MBLBP operator. MBLBP Operator consistently performs much better than the other LBP based methods for both face identification and face verification under various conditions.