一类SVD域水印问题分析及改进算法

Analysis of Problem in SVD Domain Watermarking and Improved Algorithm

分析了一类SVD域图像水印算法存在过高虚警率的原因,并给出了相应的实验结果。在此基础上,提出了一种混合DWT和SVD的图像水印算法。算法先将载体图像划分为互不重叠的块,并对每一块进行1层DWT分解,再对低频子带进行SVD分解。采用量化的方法,将Arnold置乱处理后的水印图像嵌入到SVD分解后的奇异值之中。实验表明,提出的算法克服了传统的SVD图像水印算法存在的虚警问题,对常见的信号处理如JPEG压缩、滤波、噪声等具有较好的鲁棒性。

The reasons of high false-positive detection rate problem for a class of singular value decomposition（SVD） based image watermarking algorithms were analyzed and the corresponding results were given.On this basis,a novel watermarking algorithm based on discrete wavelet transform（DWT） and SVD was proposed.First the host image is divided into non-overlap blocks and each block is applied 1-level DWT,and then SVD on the low-frequency sub-band is made.The watermarking is preprocessed by Arnold transform and then is embedded into the singular values of the low-frequency sub-band based on quantization method.Experiments show that the proposed algorithm can overcome the high false-positive detection rate problem which exists in the traditional watermarking algorithm based on SVD and robust against common signal processing such as JPEG compression,filtering,and the adding of noise.