四川大学学报（自然科学版）2025，Vol.62Issue(2)：412-424,13.DOI:10.19907/j.0490-6756.240063
基于ASIFT-Kmeans-九域卷积小波变换的抗旋转攻击图像水印算法
Watermarking algorithm based on ASIFT-Kmeans-Nine-Region convolutional wavelet transform for rotation-resilient images
摘要
Abstract
In response to the problem of poor resistance to rotation attacks in most robust digital image water-marking algorithms,this paper proposes a novel robust watermarking algorithm based on ASIFT-Kmeans-Nine-Region Convolutional Wavelet Transform.Firstly,the ASIFT method is employed to robustly extract image keypoints with well-maintained affine invariance.These feature points are then clustered into three clusters using the Kmeans algorithm,constructing stable and non-overlapping feature regions in each cluster as watermark embedding regions.Subsequently,the Nine-Region Convolutional Wavelet Transform defined in this paper is utilized to transform the image into nine different frequency domains,comprehensively captur-ing detailed information of the image to identify the most suitable watermark embedding domain.Finally,the watermark information is embedded using the DCT-SVD technique.Simulation experiments demonstrate that the proposed algorithm exhibits excellent robustness against rotation attacks and rotation combination at-tacks,as well as strong robustness against common attacks.With a Peak Signal-to-Noise Ratio(PSNR)higher than 37 dB,the normalized coefficients of the extracted blind watermark are all above 0.87.关键词
特征区域/盲水印/九域卷积小波变换/随机密钥/K-MeansKey words
Feature region/Blind watermark/Nine-domain convolution wavelet transform/Random key/K-Means分类
信息技术与安全科学引用本文复制引用
白健平,杜庆治,龙华,邵玉斌..基于ASIFT-Kmeans-九域卷积小波变换的抗旋转攻击图像水印算法[J].四川大学学报（自然科学版）,2025,62(2):412-424,13.基金项目
云南省媒体融合重点实验室开放项目(220235205) （220235205）
