Scholarlink

<p>With the rapid development of deep learning methods, their applications in various fields have gradually emerged. In the reversible data hiding (RDH) research field, the design of the prediction method is crucial for improving the performance of RDH algorithms. Currently, the Convolutional Neural Network (CNN) is introduced into RDH as a predictor because of its global field of view. Most existing RDH algorithms rely on a single predictor to generate prediction errors. However, it has been observed that exploiting different predictors based on different local textures might achieve better prediction performance. For example, the traditional rhombus predictor (RP) may be more suitable for smooth regions, while complex texture regions may be more appropriately exploited by the CNN-based predictor (CNNP). Therefore, in this paper, a hybrid prediction scheme is proposed that adaptively chooses different predictors based on the complexity measurement to achieve better performance. Experimental results demonstrate that our algorithm can obtain better performance than the state-of-the-art works.</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

CNN-Based Hybrid Prediction Scheme for Image Reversible Data Hiding

  • 期刊论文

  • 发表日期：

    2025年9月24日

查看全文
  • Yi Zhao,
  • Yi Peng,
  • Yixiang Fang,
  • Tianzhu Zhang,
  • Junxiang Wang

摘要

With the rapid development of deep learning methods, their applications in various fields have gradually emerged. In the reversible data hiding (RDH) research field, the design of the prediction method is crucial for improving the performance of RDH algorithms. Currently, the Convolutional Neural Network (CNN) is introduced into RDH as a predictor because of its global field of view. Most existing RDH algorithms rely on a single predictor to generate prediction errors. However, it has been observed that exploiting different predictors based on different local textures might achieve better prediction performance. For example, the traditional rhombus predictor (RP) may be more suitable for smooth regions, while complex texture regions may be more appropriately exploited by the CNN-based predictor (CNNP). Therefore, in this paper, a hybrid prediction scheme is proposed that adaptively chooses different predictors based on the complexity measurement to achieve better performance. Experimental results demonstrate that our algorithm can obtain better performance than the state-of-the-art works.