灰狼算法优化BP神经网络的图像去模糊复原

Image deblurring restoration of BP neural network based on grey wolf algorithm

为了解决传统复原算法在退化图像复原过程中过度依赖先验知识弊端,提出利用BP神经网络学习和泛化能力强的优点进行退化图像复原研究。首先,采用BP神经网络进行退化图像的复原研究。然后,针对BP神经网络在学习过程中由于对网络初始值的过度依赖导致网络收敛速度慢、易于陷入局部极小值的缺点。提出利用灰狼优化算法的全局搜索能力对BP神经网络的初始参数进行优化,并利用改进收敛因子与动态权重指导种群移动的方式对灰狼算法进行改进。实验表明,本文提出的改进灰狼算法优化BP神经网络复原方法与维纳滤波算法、L-R复原算法、BP神经网络和PSO-BP神经网络等复原方式相比,收敛速度和复原精度方面得到大幅度提高,在客观的评价标准结构相似度与峰值信噪比方面都获得较好的数值结果。

In order to solve the drawbacks of traditional restoration algorithms relying on a priori knowledge in the process of degraded image restoration,the advantages of BP neural network learning and generalization ability is proposed to carry out degraded image restoration research.Firstly,the BP neural network is used to recover the degraded image.Then,for the BP neural network,due to the excessive dependence on the initial value of the network,the network convergence speed is slow and it is easy to fall into the local minimum value.The global search ability of the grey wolf optimization algorithm is used to optimize the initial parameters of the BP neural network,and the gray wolf algorithm is improved by improving the convergence factor and dynamic weight to guide the population movement.The experiment shows that the improved grey wolf algorithm optimization BP neural network restoration method proposed in this paper has a high convergence speed and recovery accuracy compared with the Wiener filtering algorithm,LR restoration algorithm,BP neural network and PSO-BP neural network.It has good results in the objective evaluation standard structure similarity and peak signal to noise ratio.