基于多域变换的视频水印嵌入算法研究

Research on Video Watermark Embedding Algorithm Based on Multi-Domain Transforms

摘要: 为提升视频水印在版权保护中的实际应用效果，本文提出一种基于多域变换的视频水印嵌入算法。该算法结合离散小波变换(Discrete Wavelet Transform, DWT)与离散余弦变换(Discrete Cosine Transformation, DCT)，利用其在时频域分析与能量集中特性上的优势，实现水印信息在频域中的鲁棒嵌入。水印嵌入前通过Arnold置乱增强安全性，在帧图像中选取HL子带进行DWT分解，再对其进行DCT处理，将水印嵌入中频系数区域。实验分别在高斯噪声、椒盐噪声、剪切攻击、滤波攻击等场景下对算法的鲁棒性进行评估。结果表明，该算法在保证水印不可见性的同时，具备较强的抗攻击能力，适用于数字视频的版权保护场景。

Abstract: To enhance the practical application effect of video watermarks in copyright protection, this paper proposes a video watermark embedding algorithm based on multi-domain transforms. The algorithm integrates Discrete Wavelet Transform (DWT) and Discrete Cosine Transform (DCT) and leverages its advantages in time-frequency domain analysis and energy concentration characteristics to realize robust embedding of watermark information in the frequency domain. The watermark is pre-encrypted using Arnold scrambling to improve security, and the HL subbands are selected in the frame image for DWT decomposition, and then DCT is performed to embed the watermark into the mid-frequency coefficient region. Experiments are conducted to evaluate the robustness of the algorithm under the scenarios of Gaussian noise, salt-and-pepper noise, cropping attack, and filtering attack, respectively. The results show that the proposed method has strong anti-attack ability while ensuring the invisibility of the watermark, and is suitable for copyright protection scenarios of digital videos.