Robust image hashing techniques have been extensively studied in the recent years, with constant improvements. The challenge is that the hashing schemes need to be robust to most content-preservation operations (CPOs) and at the same time have reliable discriminative capability. One of the most important application of such image hashing techniques is image authentication. Most of the existing image hashing techniques are not robust to geometric attacks e.g. rotation and composite rotation-scaling-translation (RST). In this paper, we propose a novel image authentication model that uses a blind approach for geometric distortion correction. The hashing scheme is based on concentric square partition and statistical feature vector distances, thus making it perceptually robust and at the same time sensitive to visually distinct images. Experimental results show the proposed method outperforms some state-of-the-art models and gives desirable result for all CPOs and geometric attacks as well.