数字水印技术中,在提高水印系统的鲁棒性及有效地控制检测时的虚警概率具有十分重要的实际意义.通过对数字水印系统的研究,提出了一种新颖的基于小波变换的数字水印检测算法,该算法能够基于随机统计特征确定响应阈值,从而定量地控制检测器的虚警概率,对数字水印投入应用具有重要意义;改进了JND(just noriced difference)门限值的计算模型,该模型有效地提高了JND门限值的鲁棒性.即使在图像受到较大幅度改动的情况下,仍能很好地定位嵌入到图像中的数字水印.
In a watermarking system, improving the robustness of the watermarking system and limiting the probability of false alarms in extraction is of great practical significance. Watermarking systems were studied and several improvements proposed. The first is a novel blind watermark detection algorithm based on wavelet transform which determines alarm thresholds based on statistical characteristics. In this way the probability of false alarms can be controlled quantitatively. Moreover, the model for just noticed difference (JND) computation was also improved, which enhances the robustness of JND. This allows most embedded watermarks to be effectively located even when the image has been altered a lot.