基于混沌粒子群优化算法的CNN图像边缘提取方法

细胞神经网络(Cellular Neural Network,CNN)具有能够高速并行计算,易于硬件实现等特点,使其在未来的图像处理方面展现出了广阔的应用前景.CNN较好地探测出图像中边缘的关键在于设计出一组较好的模板参数.提出一种基于混沌粒子群优化算法求解模板参数的方法,一方面,避免了分析细胞神经网络动态性能的一系列复杂过程;另一方面,通过将搜索过程映射为对混沌轨道的遍历过程,可以使得搜索过程避免陷入局部极小,并且在模板参数的范围内能快速找到最优模板值.仿真实验表明,利用该方法设计出来的CNN去探测图像中的边缘比已有结果和利用几种经典边缘提取算子得到的边缘结果更加精确.

Image Edge Detection by Combining Chaotic Particle Swarm Optimization Algorithm and CNN

The characteristics of high-speed parallel computing and easy implementation in hardware make cellular neural network(CNN) promising in the near future.The key to better detect the edge of the image is to design a group of good templates parameters.By combining chaotic particle swarm optimization algorithm and CNN,this paper presents a method to solve the corresponding template parameters,which is very important to use the CNN to detect the edge in image.On the one hand,this method can avoid the analysis of the dynamic properties of CNN with a complex process.On the other hand,through the search process for the chaotic orbit corresponding to the traversal process,we can make the search process have the ability to avoid local minimum,and quickly find the parameters in the template.Simulation results show that the performance of edge detected by proposed method is more accurate than that by using existing methods regarding to CNN and other classical edge detection algorithms.