Robust object-based watermarking using SURF feature matching and DFT domain

In this paper we propose a robust object-based watermarking method, in which the watermark is embedded into the middle frequencies band of the Discrete Fourier Transform (DFT) magnitude of the selected object region, altogether with the Speeded Up Robust Feature (SURF) algorithm to allow the correct watermark detection, even if the watermarked image has been distorted. To recognize the selected object region after geometric distortions, during the embedding process the SURF features are estimated and stored in advance to be used during the detection process. In the detection stage, the SURF features of the distorted image are estimated and matched with the stored ones. From the matching result, SURF features are used to compute the Affine-transformation parameters and the object region is recovered. The quality of the watermarked image is measured using the Peak Signal to Noise Ratio (PSNR), Structural Similarity Index (SSIM) and the Visual Information Fidelity (VIF). The experimental results show the proposed method provides robustness against several geometric distortions, signal processing operations and combined distortions. The receiver operating characteristics (ROC) curves also show the desirable detection performance of the proposed method. The comparison with a previously reported methods based on different techniques is also provided.