Abstract
In this paper, multi-scale bandlet and local binary pattern (LBP) based method for gender recognition from faces is proposed. Bandlet is one of the multi-resolution techniques that can adapt the orientation of the edges of the face images, and thereby can better capture the texture of a face image. After extracting bandlet coefficients from face images at different scales, LBP is applied to create a histogram, which is used as the feature to a minimum distance classifier. The experiments are performed using FERET grayscale face database, and the highest accuracy of 99.13% is obtained with the proposed method.