Impact of LBP Topologies as Texture Descriptors on Ethnicity Identification

Abstract

Many ethnicity identification techniques have been developed during the past years but the problem remains are the way of using these techniques, especially local binary pattern (LBP) method is one of these techniques which has shown its superiority in ethnicity identification. The original LBP operator mainly thresholds pixels in a specific predetermined window based on the gray value of the central pixel of that window. In this work, we comparativelystudy five different configuration neighborhood topology including circle, ellipse, parabola, hyperbola, and Archimedean topology. In the ellipse topology we used eight number of neighborhood pixels with different angle (0o, 45o, 90o, and135o), also in the circle topology we used vary the number of neighborhood pixels P: P = 8, P=10, and P = 12. K-nearest neighbor (KNN) has been used for identification task.A series of experimentations hasbeen performed on1200 face images were obtained from a collection of some standard databases. The topology computations that provide highly accurate identification consist of circle, and ellipse topology. In addition, the experimental results also indicate that a good accuracy and demonstrate by increasing the number of neighborhood pixel the result will be increase.