基于显著性检测与混沌对称折叠的图像选择性加密算法

Image Selective Encryption Algorithm Based on Saliency Detection and Chaotic Symmetry Folding

当前图像选择性加密技术主要依赖明文的边缘信息实现局部像素的混淆与扩散,易泄露重要目标的形状信息,而且整个加密过程中忽略了初始明文特性,使其安全性不理想,为了解决以上的问题,提出了显著性检测耦合混沌对称折叠的图像选择性加密技术。首先,引入Ripplet变换,对输入明文完成多方向、多尺度分解,形成Ripplet系数,通过逆Ripplet变换,得到明文对应的特征图;通过高斯概率密度与信息熵来计算特征图的全局与局部显著图;再组合全局与局部显著图,根据区域生长法,检测明文的视觉显著性区域;随后,借助明文信息迭代Logistic映射,获取一组混沌数组,以定义索引扰乱机制,改变显著区域内的像素位置;最后,联合Logistic映射与对称连续折叠机制,设计了混沌折叠扩散方法,从多个方向利用不同的扩散函数对置乱后的显著性区域完成加密。实验结果显示:与已有的选择性加密方案相比,本文方法呈现出更好的安全性与密钥敏感性。

In order to solve the defect as low security caused by leaking the shape information of the important target which relying on the edge information of the plaintext to realize the confusion and diffusion of the local pixels,as well as ignoring the initial plaintext characteristics during the whole process in current image selective encryption technology,an image selective encryption algorithm based on saliency detection and chaotic symmetry folding was proposed.Firstly,the Ripplet transformation was used to decompose the input plaintext by multi-directional and multi-scale decomposition for forming Ripplet coefficients,and the corresponding feature map was obtained by inverse Ripplet transform.The global and local saliency map of feature map was calculated by Gauss probability density and information entropy.Then the global and local saliency map is combined to extract the visual saliency area of the plaintext image.The visual saliency area of plaintext was detected based on the region growing method and combining the global and local saliency map.Then the index confusion mechanism was defined according to the random sequence generated by using the plaintext pixel to iterate logistic mapping for scrambling the pixels in the region of interest.Finally,a chaotic folding diffusion method was designed by combining Logistic map and symmetric continuous folding mechanism to encrypt the visual saliency region from many directions by using different diffusion functions.The experimental results show that this algorithm has higher security and better key sensitivity compared with the existing image encryption technology.