C3MW: A Novel Comprehensive-Monitoring-Motivated Multi-model Watermarking Scheme for Tamper Detection and Self-recovery
Abstract
In this paper, we propose a novel Comprehensive-Monitoring-Motivated Multi-model Watermarking (C3MW) Scheme for tampering region localization and self-recovery for 4K images. To generate the Comprehensive-Monitoring-Motivated Multi-model (C3M) Watermark, the MultiModel Authentication Bits Generation (MMAG) method and the Adaptive Block SignificanceBased Recovery Bits Generation (ASRG) method, are proposed. The MMAG aims to monitor the various bit-plane information for detecting the possible tampering in a more comprehensive manner. When performing image tampering detection, once one of the multiple models is triggered, the corresponding parcel will be marked as tampered. On basis of the detected regions, we propose a fine-tuning-based image recovery method, where the extracted recovery data consist of the fused Adaptive Block Significance (ABS) and bitmaps, while the calculated recovery data consist of the watermark information which is calculated from the received image. We conduct experiments on two public databases, respectively, the BOWS2 Dataset and the LIU4K-v2 Dataset. Comparisons with existing state-of-the-art works have been performed on the BOWS2 dataset, and our scheme improves the precision and F1 Score by 7.27% and 3.30%, respectively. It is clear from these results that our method has a better performance than others.
Qiyuan Zhang
Phd student (since 2021.8)
