Computer and soft Engineer Department, Anhui Institute of Information Technology, Wuhu, Anhui, China.
Shanghai DataSeed Information Technology, Shanghai, China.
PLoS One. 2024 Feb 2;19(2):e0297632. doi: 10.1371/journal.pone.0297632. eCollection 2024.
Numerous image authentication techniques have been devised to address the potential security issue of malicious tampering with image content since digital images can be easily duplicated, modified, transformed and diffused via the Internet transmission. However, the existing works still remain many shortcomings in terms of the recovery incapability and detection accuracy with extensive tampering. To improve the performance of tamper detection and image recovery, we present a block mapping and dual-matrix-based watermarking scheme for image authentication with self-recovery capability in this paper. The to-be-embedded watermark information is composed of the authentication data and recovery data. The Authentication Feature Composition Calculation algorithm is proposed to generate the authentication data for image tamper detection and localization. Furthermore, the recovery data for tampered region recovery is comprised of self-recovery bits and mapped-recovery bits. The Set Partition in Hierarchical Trees encoding algorithm is applied to obtain the self-recovery bits, whereas the Rehashing Model-based Block Mapping algorithm is proposed to obtain the mapped-recovery bits for retrieving the damaged codes caused by tampering. Subsequently, the watermark information is embedded into the original image as digital watermarking with the guidance of a dual-matrix. The experimental results demonstrate that comparing with other state-of-the-art works, our proposed scheme not only improves the performance in recovery, but also extends the limitation of tampering rate up to 90%. Furthermore, it obtains a desirable image quality above 40 dB, large watermark payload up to 3.169 bpp, and the effective resistance to malicious attack, such as copy-move and collage attacks.
已经设计了许多图像认证技术来解决数字图像可以通过互联网传输轻松复制、修改、转换和传播的潜在安全问题。然而，现有的工作在广泛篡改的恢复能力和检测精度方面仍然存在许多缺点。为了提高篡改检测和图像恢复的性能，我们在本文中提出了一种具有自恢复能力的基于块映射和双矩阵的图像认证水印方案。待嵌入的水印信息由认证数据和恢复数据组成。提出了认证特征合成计算算法来生成用于图像篡改检测和定位的认证数据。此外，用于篡改区域恢复的恢复数据由自恢复位和映射恢复位组成。应用分层树集合划分编码算法获取自恢复位，提出基于再散列模型的块映射算法获取映射恢复位，以恢复篡改引起的损坏代码。随后，在双矩阵的指导下，将水印信息嵌入原始图像作为数字水印。实验结果表明，与其他最先进的工作相比，我们提出的方案不仅提高了恢复性能，而且将篡改率的限制扩展到 90%。此外，它获得了超过 40dB 的理想图像质量、高达 3.169 bpp 的大水印负载，以及对恶意攻击（如复制粘贴和拼贴攻击）的有效抵抗。
