基于MEA-WNN神经网络的图像复原方法

Image Restoration Based on Wavelet Neural Network Optimized by Mind Evolutionary Algorithm

摘要: 由于小波神经网络图像还原的效果一定程度上受初始值的影响，因此本文提出了一种基于思维进化算法(Mind Evolutionary Algorithm, MEA)优化小波神经网络方法。思维进化算法本身具有很强的全局搜索能力，因此先用MEA方法得到小波神经网络的初始值，再训练小波神经网络。实验证明，与BP、思维进化算法的BP神经网络(MEA-BP)相比，MEA-WNN方法复原的图像获得了更好的结果。

Abstract: Since the effect of wavelet neural network image restoration is largely affected by the initial value and weight, a wavelet neural network optimization method based on mind evolutionary algorithm (MEA) is proposed. Thinking evolution itself has strong global search ability. Therefore, before training wavelet neural network, mind evolution algorithm is used to obtain the initial value and weight of the network. Experiments show that compared with BP, BP neural network (MEA-BP) based on mind evolutionary algorithm, the image restored by MEA-WNN method obtains better results.