Scholarlink

Purpose <p>Medical images stored in the DICOM format are highly vulnerable once they leave secure healthcare environments, exposing sensitive data to unauthorized manipulation and ownership disputes. Therefore, this work proposes a dual watermarking scheme specifically designed to protect high bit-depth grayscale medical images by providing both copyright protection and tamper localization.</p> Methods <p>The proposed approach embeds two watermarks: a robust watermark based on convolutional coding and a semi-fragile watermark that differentiates between benign DICOM compression and malicious tampering. The spatial domain embedding method distributes both watermarks across non-overlapped 12 × 12 image blocks. Additionally, an interpolation technique is applied to improve the perceptual quality of the image when no tampering is detected.</p> Results <p>Experiments on tomography, radiography, and X-ray fluoroscopy datasets with native resolutions and bit depths show competitive image quality, with mean PSNR values of 51.97 dB, 55.34 dB, and 41.76 dB, respectively. These values increase to 65.53 dB, 57.53 dB, and 45.44 dB after interpolation. The robust watermark allows the optimal recovery of an identification logo even with up to 25% attacked area and recognizable recovery up to 40%. The semi-fragile watermark accurately detects tampering with a mean error of 1%, while tolerating DICOM-standard compressions.</p> Conclusions <p>The proposed method achieves competitive performance in both imperceptibility and robustness compared with state-of-the-art schemes, while preserving the native DICOM format. Its spatial-based embedding simplifies implementation and enables practical protection of medical images without extensive preprocessing or external models.</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

Copyright protection and tamper detection in medical images via a dual watermark scheme

  • 期刊论文

  • 发表日期：

    2026年2月18日

查看全文
  • Eduardo Fragoso-Navarro,
  • Rodrigo Eduardo Arevalo-Ancona,
  • Manuel Cedillo-Hernandez,
  • Francisco Javier Garcia-Ugalde

摘要

Purpose

Medical images stored in the DICOM format are highly vulnerable once they leave secure healthcare environments, exposing sensitive data to unauthorized manipulation and ownership disputes. Therefore, this work proposes a dual watermarking scheme specifically designed to protect high bit-depth grayscale medical images by providing both copyright protection and tamper localization.

Methods

The proposed approach embeds two watermarks: a robust watermark based on convolutional coding and a semi-fragile watermark that differentiates between benign DICOM compression and malicious tampering. The spatial domain embedding method distributes both watermarks across non-overlapped 12 × 12 image blocks. Additionally, an interpolation technique is applied to improve the perceptual quality of the image when no tampering is detected.

Results

Experiments on tomography, radiography, and X-ray fluoroscopy datasets with native resolutions and bit depths show competitive image quality, with mean PSNR values of 51.97 dB, 55.34 dB, and 41.76 dB, respectively. These values increase to 65.53 dB, 57.53 dB, and 45.44 dB after interpolation. The robust watermark allows the optimal recovery of an identification logo even with up to 25% attacked area and recognizable recovery up to 40%. The semi-fragile watermark accurately detects tampering with a mean error of 1%, while tolerating DICOM-standard compressions.

Conclusions

The proposed method achieves competitive performance in both imperceptibility and robustness compared with state-of-the-art schemes, while preserving the native DICOM format. Its spatial-based embedding simplifies implementation and enables practical protection of medical images without extensive preprocessing or external models.