提出了一种可用于版权认证的基于小波变换域自适应量化步长的图像盲水印算法.该算法首先在DWT（离散小波变换）的低频子带中选择感知重要系数作为水印的嵌入区间,然后选择一个同时可满足不可见性与鲁棒性要求的最小量化步长,对所选择的重要系数进行量化,再通过自适应增大量化步长,把二值水印信息交替地嵌入到不同的量化层上.作为这种算法的一个特例,就是用所选系数的低5位表示水印信息.在嵌入过程中,根据小波变换所具有的良好的空域局部化特性,采用了分块舍入技术,实现了水印信息的精确嵌入;在检测过程中,提出了一种新颖的差分检测方法,增强了系统的鲁棒性.实验结果证明,该算法具有嵌入信息量大,水印透明性好,对各种常见的图像处理操作、有损压缩、裁剪和几何变换抵抗性强等特点.

A new image blind watermarking algorithm based on adaptive quantization step in discrete wavelet transform(DWT) is proposed,which can be used in copyright authentication.In the proposed algorithm,the perceptually significant coefficients are selected from LL subband to be the embedding domain,and a minimum quantization step,which guarantees the invisibility and robustness,is used to quantify the selected coefficients.The binary watermarks are alternately embedded on the different quantified levels by adaptively increasing the quantization step.As a special example,the least 5 bits of the selected coefficients are utilized to express the watermark.Considering the localization character of wavelet transform in spatial domain,a block-based rounding method is adopted to achieve the purpose of precise embedding.As to the detecting algorithm,a novel difference detecting algorithm is created to improve the robustness of system.The experimental results show that the proposed algorithm has the following characters: large information capacity,invisibility and robustness against attacks such as some common image processing operations,lossy compression,cropping and some geometric transformations etc.