在充分考虑人类视觉系统掩蔽特性的基础上,首先给出基于图像分块的临界噪声阈值矩阵JND（justnoticeable difference）,进而提出了一种结合离散小波变换（DWT）和离散余弦变换（DCT）域的采用灰度图像作为水印的具有较好鲁棒性的公开水印技术.实验结果表明,该算法对常见图像处理操作、有损压缩和裁剪、挤压、像素移位等几何变换有较好的鲁棒性,特别是信号增强操作处理几乎不影响水印的正确提取.

This paper is based on taking the masking characteristics of the human visual system into account.First,we give the method that can calculate the just noticeable difference(JND) matrix.Secondly,we propose a public watermarking of gray-level still images combining DWT and DCT.The results of experiment indicate that this watermarking scheme is much robust against noise interference,some geometric distortions and commonly used image processing techniques,especially the signal enhanced processing almost can not affect the correct extraction of watermark.