Scholarlink

<p>Reversible data embedding is a new field to enable secure image communication in many applications that require exact restoration of original content, including medical diagnosis, forensic imagery, and preservation of Cultural Heritage. The current technologies for Reversible Data Hiding (RDH), such as Histogram Shifting and Difference Expansion, contain many drawbacks: high computational complexity, large amounts of distortion as payload grows, and the need for auxiliary information (e.g., location maps) to be used. Therefore, these factors limit the application of conventional RDH techniques in real-time or resource constrained environments. To address these shortcomings, we present our proposed Adaptive XOR-Prediction (AXP-RDH) framework that provides an exceptionally lightweight RDH method with very high embedding capacity, strict reversibility, and very little arithmetic overhead. Our new approach uses the bitwise XOR of a stretcher’s output to generate a causal predictor, then uses this predictor with an adaptive prediction-error expansion strategy to determine how to expand a payload. Our proposed approach does not require a global histogram during payload expansion and does not require the use of floating-point arithmetic. Additionally, it does not require any auxiliary information. The processes of embedding and extracting information utilize only integer arithmetic operations and bitwise logical operations, providing high performance for use on constraint platforms such as edge computing devices. Extensive experiments conducted on medical MRI images of brain tumours, along with images from a world heritage repository and the grayscale image dataset, show that the proposed method is capable of achieving complete recovery while maintaining high payload capacity, PSNR and SSIM values compared to traditional methods of encrypting information with visual cryptography and, furthermore, outperforming representative traditional methods in terms of lightweight operation and attractive performance levels. Finally, both the security and robustness analyses indicate that the proposed system has the potential to be incorporated into future generations of secure 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

Adaptive XOR-based causal prediction for lightweight and fully reversible data hiding

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月15日

查看全文
  • Tarak Hussain,
  • Faizal Nujumudeen,
  • Zulfikar Ali Ansari,
  • Hemlata Pant,
  • Mohammadi Begum Jeelani,
  • Gyan Chand Yadev

摘要

Reversible data embedding is a new field to enable secure image communication in many applications that require exact restoration of original content, including medical diagnosis, forensic imagery, and preservation of Cultural Heritage. The current technologies for Reversible Data Hiding (RDH), such as Histogram Shifting and Difference Expansion, contain many drawbacks: high computational complexity, large amounts of distortion as payload grows, and the need for auxiliary information (e.g., location maps) to be used. Therefore, these factors limit the application of conventional RDH techniques in real-time or resource constrained environments. To address these shortcomings, we present our proposed Adaptive XOR-Prediction (AXP-RDH) framework that provides an exceptionally lightweight RDH method with very high embedding capacity, strict reversibility, and very little arithmetic overhead. Our new approach uses the bitwise XOR of a stretcher’s output to generate a causal predictor, then uses this predictor with an adaptive prediction-error expansion strategy to determine how to expand a payload. Our proposed approach does not require a global histogram during payload expansion and does not require the use of floating-point arithmetic. Additionally, it does not require any auxiliary information. The processes of embedding and extracting information utilize only integer arithmetic operations and bitwise logical operations, providing high performance for use on constraint platforms such as edge computing devices. Extensive experiments conducted on medical MRI images of brain tumours, along with images from a world heritage repository and the grayscale image dataset, show that the proposed method is capable of achieving complete recovery while maintaining high payload capacity, PSNR and SSIM values compared to traditional methods of encrypting information with visual cryptography and, furthermore, outperforming representative traditional methods in terms of lightweight operation and attractive performance levels. Finally, both the security and robustness analyses indicate that the proposed system has the potential to be incorporated into future generations of secure imaging systems.