A semi-fragile watermarking tamper localization method based on QDFT and multi-view fusion

Accurate tamper localization is one of the most challenging problems in semi-fragile watermarking authentication schemes. The existing semi-fragile watermarking methods have some problems, such as blurred localization shape and high false alarm rate, which stem from the rough analysis of tampering information. To address these problems, this paper proposes a semi-fragile watermarking tamper localization method based on quaternion discrete Fourier transform (QDFT) and multi-view fusion, which embeds two watermarks in the QDFT domain for robust resistance to the geometric attack and integrity protection of host image, respectively. On the receiver side, considering that the difference map in binary mode is difficult to identify tampering, the global information and local smoothing cues are used in the method, to obtain the multi-scale candidate maps with real values instead of binary. Then, local adaptive fusion is accomplished by minimizing the energy function, to obtain a more consistent single tampering map. The design of the energy function integrates the image content and multi-scale feature views. Finally, a tamper seed-based propagation strategy is designed to generate a binary map of the tampered regions. Experimental results show that the proposed method provides better resistant to signal processing attacks, and the F-measure value of tamper localization is improved by 11.17% on average compared to the state-of-the-art methods.