In this paper, a novel local image representation method based on Walsh-Hadamard Transform (WHT) called Local WHT (LWHT) is introduced. It is explained how to create complex images from outputs of LWHT which are real valued images. The proposed method is improved by application of YWHD method to landmarks and by cascaded YWHD method. Experiments on FERET dataset show that YWHD produce good results among other local image representation methods like LBP and LGBP and improve WHT based face recognition systems.