Robust face recognition under uncontrolled illumination conditions is one of the key challenges for real-time face recognition systems. Weber-face (WF) is an illumination insensitive face representation based on Weber' law. In this letter, we develop a generalized Weber-face (GWF) which extracts the statistics of multi-scale information from face images. By assigning different weights to the inner-ground and outer-ground we further develop a weighted GWF (wGWF) version. Based on our experiments on the extended Yale-B and FERET face database we show that the proposed methods are robust to illumination variations and can obtain promising performance comparable with existing approaches. © 2014 Elsevier B.V.