基于小波压缩深度学习重构的多图像光学加密  

Multi-image optical encryption based on wavelet compression and deep learning reconstruction

作　　者：郭媛[1] 贾德宝 敬世伟 翟平 GUO Yuan;JIA De-bao;JING Shi-wei;ZHAI Ping(College of Computer and Engineering,Qiqihar University,Qiqihar 161006,China)

机构地区：[1]齐齐哈尔大学计算机与工程学院,黑龙江齐齐哈尔161006

出　　处：《计算机工程与设计》2024年第2期367-375,共9页Computer Engineering and Design

基　　金：国家自然科学基金项目(61872204);黑龙江省自然科学基金项目(LH2021F056);黑龙江省省属高等学校基本科研业务费科研基金项目(135509113);齐齐哈尔大学研究生创新科研基金项目(YJSCX2022049)。

摘　　要：为解决多图像加密算法密文体积大、加密效果差、重构效果不理想等问题,提出一种基于小波压缩和深度学习重构的多图像光学加密方法。利用小波压缩提取多图像的低频部分,将重排后的新图放入改进的FDT-DRPE光学加密系统中得到密文;利用矢量分解和螺旋相位变换克服FDT-DRPE不敏感问题;构造的L_S混沌系统提高密钥敏感性。提出新型深度学习网络模型BHCN,解决传统图像重构精度不高问题。实验结果表明,密文体积可压缩至原图的1/4,重构图像的峰值信噪比为34.57 dB,结构相似性为0.9521,与同类文献相比,速度更快,重构效果更好,安全性更高。A multi-image optical encryption method based on wavelet compression and deep learning reconstruction was proposed to solve the problems such as large ciphertext,poor encryption effect and unsatisfactory reconstruction effect.The wavelet compression was used to extract the low-frequency parts of multiple images,and the new images were put into the improved FDT-DRPE optical encryption system to get the ciphertext.The insensitivity of FDT-DRPE was overcome by vector decomposition and helical phase transformation.The constructed L_S chaotic system improved the key sensitivity.A deep learning network model BHCN was proposed to solve the problem of low accuracy of traditional image reconstruction.Experimental results show that the volume of ciphertext can be compressed to 1/4 of the original image,the peak signal-to-noise ratio of reconstructed image is 34.57 dB,and the structural similarity is 0.9521.Compared with similar literatures,the reconstruction speed is higher,the reconstruction effect is better,and the security is higher.

关 键 词：多图像光学加密 深度学习 小波压缩 菲涅尔双随机相位编码 矢量分解 混沌系统 比特分层 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]