基于正交基函数神经网络的图像加密算法仿真  

Simulation of Image Encryption Algorithm Based on Orthogonal Basis Function Neural Network

作　　者：林青[1] 戴慧珺[2] 马文涛[3] 

机构地区：[1]西安培华学院电器信息工程学院,陕西西安710125 [2]西安交通大学电信学院,陕西西安710049 [3]西安航空学院基础部,陕西西安710077

出　　处：《计算机仿真》2013年第10期416-421,共6页Computer Simulation

基　　金：国家自然科学基金(61071126)

摘　　要：研究图像加密优化问题,传统混沌系统对初始条件和参数具有很强的敏感性,图像加密中常用的混沌系统的另外一些特征,如同步性、参数少等,也是导致不安全的因素。为了克服混沌系统的缺陷,利用正交基函数神经网络具有简单易实现,且可以建立输入输出之间的非线性关系的特性,构建正交基函数神经网络,以三种混沌模型产生的混沌序列构造网络的输入输出,对网络进行训练,然后再使用新的混沌序列,结合训练得到的权值来获得网络的输出,建立密钥序列。采用一种可逆的象素值替代变换,根据密钥序列对彩色图像的各分量进行象素值替代,重构得到加密图像。仿真结果表明,改进算法具有较强的密钥空间和很强的密钥敏感性,是一种安全有效的加密方法。The chaotic system is highly sensitive for its initial conditions and parameters, therefore it has been widely used in image encryption. However, several other features of chaotic system, such as synchronization and less parameters, can result in unsafe factors. In order to overcome the defects of chaotic system, for the orthogonal basis function neural network has the characteristic of simple and easy to realize and can established the non - linear rela- tionship, the orthogonal basis function neural network was constructed. Taking chaotic sequences generated by three kinds of chaotic models as the input and output, we trained the neural network with the input and output, then used the new chaotic sequence combined with trained weights to obtain the network output and establish key sequences. A reversible pixel replacement transform was proposed, according to the sequence of keys, the pixel of each component of the color image was substituted and the encrypted image was obtained by reconstruction. The simulation results show that the proposed algorithm has a powerful key space and strong key sensitivity, and it is a safe and effective en- cryption method.

关 键 词：图像加密 正交基函数 正交基 神经网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]