DCT域基于α稳定分布的数字图像水印检测器

Digital Image Watermark Detector Design Based on α Distribution in DCT Domain

提出了离散余弦变换(DCT)域基于α稳定分布的分数低阶统计量水印检测器设计的理论推导和实验证明。理论分析和实验结果表明:传统的DCT域水印检测是将变换系数的交流分量作二阶统计量Gausissian分布假设,并依此作为检测器设计依据,非二阶统计量检测范围与精度受到限制与影响;基于α稳定分布的假设更适合数字图像水印处理的实际应用,提案方法的约束条件更为宽泛,与对照方案相比,水印提取失真率降低了7.4%,且适用对象更具一般性。

Based on the hypothesis of a stable distribution,the paper proposes a new method to design watermark detectors in discrete cosine transform（DCT） domain. By the means of fractional lower order statistics estimation, the method improves the traditional detector which is based on Gaussian distribution hypothesis. In accordance with this method,2nd order statistics theory are limited in actual processing. The hypothesis of a stable distribution is more suitable for practical processing of digital image watermark. The theoretical analysis and experimental results show that proposed method is more efficient than tradition one and it has wider applications.