| 梁梦华,赵鸿图.基于混沌映射和可逆神经网络的图像隐写算法研究[J].高技术通讯(中文),2026,36(2):147~156 |
| 基于混沌映射和可逆神经网络的图像隐写算法研究 |
| Research on image steganography algorithm based on chaotic mapping and invertible neural network |
| |
| DOI：10. 3772 / j. issn. 1002 - 0470. 2026. 02. 004 |
| 中文关键词: 图像隐写； 混沌映射； 可逆神经网络； 图像质量； 抗隐写分析； 鲁棒性 |
| 英文关键词: image steganography, chaotic mapping, invertible neural network, image quality, anti-steganography analysis, robustness |
| 基金项目: |
| 作者 | 单位 | | 梁梦华 | （河南理工大学物理与电子信息学院焦作 454001） | | 赵鸿图 | |
|
| 摘要点击次数: 159 |
| 全文下载次数: 130 |
| 中文摘要: |
| 针对大容量隐写规模导致的图像质量下降问题，本文设计了一种全新的可逆大容量图像隐写网络，该网络结合了混沌映射与可逆神经网络，在保证大容量隐写的同时极大提高了隐写的安全性。首先，设计了一个具有线平衡的四维混沌系统，通过混沌映射加密算法对秘密图像的内容进行置乱，从而解决了秘密图像信息在传输过程中的泄露问题。其次，对载体图像进行预处理，在Y通道隐藏秘密图像，显著增强了隐写容量。在DIV2K和COCO数据集上进行了大量实验，载体图像与载密图像的峰值信噪比(peak signal-to-noise ratio，PSNR)高达41.63dB，秘密图像与恢复图像的PSNR为43.29dB。大量的实验结果表明，在大容量隐写条件下，本文方法不仅能保持优异的隐写性能，同时能高保真度显示隐写图像，其抗隐写能力远远优于现有的先进算法。 |
| 英文摘要: |
| To address the problem of image quality degradation caused by large-scale steganography, a new reversible steganography network is designed in this paper. This network combines chaotic mapping and invertible neural network, greatly improving the security of steganography while ensuring large-scale capacity. Firstly, a four-dimensional chaotic system with line balance is designed. The contents of the secret image are scrambled using a chaotic mapping encryption algorithm to prevent secret image information leakage during transmission. In addition, the cover image is preprocessed to hide the secret image in the Y channel, significantly enhancing the steganographic capacity. Extensive experiments have been conducted on the DIV2K and COCO datasets. The peak signal-to-noise ratio (PSNR) of the cover image and stego image is as high as 41.63dB, and that of the secret image and recovered image is 43.29dB.The experimental results demonstrate that the proposed method not only maintains excellent steganographic performance under large capacity steganographic conditions but also produces steganographic images with high fidelity. Furthermore, its anti-steganographic capability is far superior to existing advanced algorithms. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
