一种过完备字典实现指纹图像的鲁棒水印算法

Robust Image Watermarking Algorithm Based on Block Compressed Sensing and Overcomplete Dictionary

针对图像的信息取证与高精度恢复要求,结合分块压缩感知理论及其过完备字典的稀疏表示,提出一种鲁棒的自适应图像水印实现算法.基于水印图像特征结构的分布特点,将其分解为平滑、纹理和边缘的同性结构区域,组成多成分过完备字典;并结合可变的采样率,构建了一种结构自适应多成分稀疏表示和水印实现模型.同时,利用过完备字典的多成分结构和稀疏表示系数的先验知识,有效实现数字水印提取和原始指纹图像重构,实验结果表明该方法具有更高的准确性和较好的鲁棒性.

The structural characteristics of overcomplete dictionary have a vital influence on the performance of image recognition,therefore,the design of overcomplete dictionary is a fundamental task in image sparse representations.With the requirements of information forensics and high precision restoration in the image,this study proposes an adaptive robust image watermarking algorithm,which consists of the block compressed sensing and the sparse representation of overcomplete dictionary.A structure adaptive multicomponent sparse representation and watermark implementation model are constructed by the size distribution features of block random projection energy of watermarking image,and are classified into smooth,texture and edge structure isotropic regions.In addition,a multi-component overcomplete dictionary is designed with consistent morphology.The extract of watermark and fingerprint image reconstruction both benefited from the prior knowledge about multi-component construction and sparse representation of dictionary.The experiment results verify the effectiveness of the proposed method.