Scholarlink

<p>The development of technology makes it easier to copy and restore digital data. More than a trillion bytes of data are created and shared online every day, and there is a major problem with the reliability of this data in its digital form. There are many forms of protection available in the Bitcoin and image watermarking communities. This study proposes an advanced digital image watermarking system that employs optimal deep learning to increase the security of information sharing. The watermark image is initially decomposed by means of a discrete cosine transform (DCT). The recursive encryption process is then applied to the watermark image, further enhancing the security of the model. The watermark image is subsequently incorporated into the cover image by employing the proposed Optimized DenseEfficientNet model. The accuracy of embedding is enhanced in this scenario through the application of the Improved Aquila Optimization (IAqO) technique to the hybrid deep learning model. In the proposed study, the Python tool is favored for simulation, while USC-SIPI datasets are utilized for analysis. According to the simulation results, the suggested model outperforms earlier techniques in terms of embedded phase SSIM (0.874), PSNR (31.68), MSE (44.08) and extracted phase SSIM (0.913), PSNR (36.63) and MSE (53.42) for USC-SIPI.</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

Optimized DenseEfficientNet based digital image watermarking with recursive encryption algorithm

  • 期刊论文

  • 发表日期：

    2026年4月12日

查看全文
  • Sambhaji Marutirao Shedole,
  • Santhi Vaithiyanathan

摘要

The development of technology makes it easier to copy and restore digital data. More than a trillion bytes of data are created and shared online every day, and there is a major problem with the reliability of this data in its digital form. There are many forms of protection available in the Bitcoin and image watermarking communities. This study proposes an advanced digital image watermarking system that employs optimal deep learning to increase the security of information sharing. The watermark image is initially decomposed by means of a discrete cosine transform (DCT). The recursive encryption process is then applied to the watermark image, further enhancing the security of the model. The watermark image is subsequently incorporated into the cover image by employing the proposed Optimized DenseEfficientNet model. The accuracy of embedding is enhanced in this scenario through the application of the Improved Aquila Optimization (IAqO) technique to the hybrid deep learning model. In the proposed study, the Python tool is favored for simulation, while USC-SIPI datasets are utilized for analysis. According to the simulation results, the suggested model outperforms earlier techniques in terms of embedded phase SSIM (0.874), PSNR (31.68), MSE (44.08) and extracted phase SSIM (0.913), PSNR (36.63) and MSE (53.42) for USC-SIPI.