中文摘要：

在前期研究工作的基础上，提出了一种不同于现有以信息论为基础的图像信息隐藏容量分析方法，该方法基于联想记忆神经网络的吸引子和吸引域理论，研究数字图像的信息可隐藏容量的极限问题，并把信息隐藏容量与最小可检测容量问题——隐藏信息量的上限和下限问题统一在同一个理论框架之内。研究结果表明，如果隐藏信息后的图像与宿主图像之间的汉明距离超出神经网络吸引域范围，就无法正确地检测出隐藏的信息；神经网络的容量决定了在一定检测错误率情况下最少需要嵌入的信息量；图像中可以隐藏的最大信息量取决于神经网络的吸引域；最小可检测信息量取决于神经网络的吸引子。该方法是对现有基于信息论方法的有益补充。

英文摘要：

On the basis of the previous research work, this paper proposes a new method for analysis of the information hiding limits in digital images based on the theories of attractors and attraction basins of neural networks. The method unifies the upper limit and lower limit of information hiding, namely the maximum information capacity and the minimum detectable information capacity in a same theory frame. The results of the research show that if the Hamming distance between the information hidden image and the original image is out of the bounds of the attraction basin, the hiding information can not be detected correctly; the capacity of neural networks decides the minimum capacity of detectable information at a certain error rate of detection; the attraction basins of neural networks decide the upper limits of information hiding; and the attractors of neural networks decide the lower information limits. The method is of complementary significance to the present ones based on the information theory.