In this paper, a robust image hashing framework is presented using discrete cosine transformation and singular value decomposition. Firstly, the input image is normalized using geometric moment and normalized coefficients are divided into non-overlapping blocks. The selected blocks based on a peace-wise non-linear chaotic map are transformed using discrete cosine transom followed by singular value decomposition. Then a feature matrix is constructed in reliance on Hessian matrix and the final hash values are obtained. The proposed hashing system is resilient to different content-preserving image distortions such as geometric and filtering operations. The simulated results demonstrate the efficiency proposed framework in terms of security and robustness. © 2018 ACM.