现有图像隐写算法没有很好地应用修改聚集原则，导致隐写算法的安全性仍然不够理想。经典隐写算法MiPOD (Minimizing the Power of Optimal Detector)拥有较高的安全性，但它使用随机修改方向的方式对像素进行修改，不符合修改聚集原则。为了提升隐写算法的安全性，在MiPOD的基础上，本文使用最小值滤波将代价矩阵局部区域内最小值的修改方式赋予整个局部，实现局部修改聚集；此外使用感知哈希算法寻找图像中相似的部分，赋予相似区域相同的修改方向，以达到整体修改聚集的目的。实验结果表明，本文算法相较于MiPOD，抵抗SRM特征的检错率平均提升了0.72%，抵抗maxSRM特征的检错率平均提升了0.61%。 Now the image steganography algorithm does not apply the clustering rule very well. It leads to the security of steganography algorithm that is still not ideal. MiPOD (Minimizing the Power of Optimal Detector) is a classical steganography algorithm, which has a high security. But it uses the method of randomly changing the direction to modify the pixels, which does not conform to the clustering rule. In order to improve the security of the steganography algorithm, our algorithm which base on MiPOD uses the minimum filter, the pixel changed method corresponding to the minimum value in the local area of the cost matrix is given to the whole local area. It can realize local clustering. In addition, the perceptual hash algorithm is used to find the similar parts of the image, and the similar regions are given the same changed direction. It can realize overall clustering. Comparing with the MiPOD, the testing error rate of SRM is improved by 0.72%, and that of maxSRMd2 is improved by 0.61%.
展开▼
机译：现有图像隐写算法没有很好地应用修改聚集原则，导致隐写算法的安全性仍然不够理想。经典隐写算法MiPOD (Minimizing the Power of Optimal Detector)拥有较高的安全性，但它使用随机修改方向的方式对像素进行修改，不符合修改聚集原则。为了提升隐写算法的安全性，在MiPOD的基础上，本文使用最小值滤波将代价矩阵局部区域内最小值的修改方式赋予整个局部，实现局部修改聚集;此外使用感知哈希算法寻找图像中相似的部分，赋予相似区域相同的修改方向，以达到整体修改聚集的目的。实验结果表明，本文算法相较于MiPOD，抵抗SRM特征的检错率平均提升了0.72%，抵抗maxSRM特征的检错率平均提升了0.61%。 Now the image steganography algorithm does not apply the clustering rule very well. It leads to the security of steganography algorithm that is still not ideal. MiPOD (Minimizing the Power of Optimal Detector) is a classical steganography algorithm, which has a high security. But it uses the method of randomly changing the direction to modify the pixels, which does not conform to the clustering rule. In order to improve the security of the steganography algorithm, our algorithm which base on MiPOD uses the minimum filter, the pixel changed method corresponding to the minimum value in the local area of the cost matrix is given to the whole local area. It can realize local clustering. In addition, the perceptual hash algorithm is used to find the similar parts of the image, and the similar regions are given the same changed direction. It can realize overall clustering. Comparing with the MiPOD, the testing error rate of SRM is improved by 0.72%, and that of maxSRMd2 is improved by 0.61%.
展开▼
