Scholarlink

<p>In this study, we propose a robust and secure medical image watermarking scheme that integrates deep learning, chaotic encryption, and advanced multiresolution analysis. The core of the method relies on a convolutional variational autoencoder (VAE) trained on a custom dataset of chaotically encrypted medical images to generate high-entropy random sequences. These sequences serve as cryptographic keys for watermark encryption using both pixel permutation and modular subtraction.The watermark, a 256 × 256 grayscale image, is then embedded into the host medical image through singular value decomposition (SVD) applied to the low-frequency (LL) sub-band of a Beta wavelet decomposition. This hybrid embedding strategy ensures both imperceptibility and robustness by leveraging the structural preservation properties of singular values and the frequency localization advantages of Beta wavelets. Experimental results on diverse medical image modalities, CT, MRI, X-ray, mammogram, and fundus—demonstrate excellent imperceptibility (PSNR = 79.32 dB), strong robustness to 25 common image attacks (NC &gt; 0.97), and high cryptographic security (NPCR = 100%, UACI = 36.05%). The proposed method significantly outperforms classical DWT-SVD and DCT-based approaches and is computationally efficient, making it highly suitable for secure watermarking in clinical imaging systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

βWave-VAE: A hybrid beta wavelet and variational autoencoder framework for secure medical image watermarking

  • 期刊论文

  • 发表日期：

    2026年6月17日

查看全文
  • Rayen Ben Salah,
  • Mourad Zaied

摘要

In this study, we propose a robust and secure medical image watermarking scheme that integrates deep learning, chaotic encryption, and advanced multiresolution analysis. The core of the method relies on a convolutional variational autoencoder (VAE) trained on a custom dataset of chaotically encrypted medical images to generate high-entropy random sequences. These sequences serve as cryptographic keys for watermark encryption using both pixel permutation and modular subtraction.The watermark, a 256 × 256 grayscale image, is then embedded into the host medical image through singular value decomposition (SVD) applied to the low-frequency (LL) sub-band of a Beta wavelet decomposition. This hybrid embedding strategy ensures both imperceptibility and robustness by leveraging the structural preservation properties of singular values and the frequency localization advantages of Beta wavelets. Experimental results on diverse medical image modalities, CT, MRI, X-ray, mammogram, and fundus—demonstrate excellent imperceptibility (PSNR = 79.32 dB), strong robustness to 25 common image attacks (NC > 0.97), and high cryptographic security (NPCR = 100%, UACI = 36.05%). The proposed method significantly outperforms classical DWT-SVD and DCT-based approaches and is computationally efficient, making it highly suitable for secure watermarking in clinical imaging systems.