基于头脑风暴优化算法的BP神经网络模糊图像复原  

BP Neural Network Fuzzy Image Restoration Based on Brain Storming Optimization Algorithm

作　　者：梁晓萍 郭振军[1] 朱昌洪 LIANG Xiaoping;GUO Zhenjun;ZHU Changhong(Guilin University of Aerospace Technology,Guilin 541004,China)

机构地区：[1]桂林航天工业学院

出　　处：《电子与信息学报》2019年第12期2980-2986,共7页Journal of Electronics & Information Technology

基　　金：2019年度广西高校中青年教师科研基础能力提升项目(2019KY0802);桂林航天工业学院电子信息重点学科及物联网与大数据应用研究中心项目(KJPT201805)~~

摘　　要：该文提出一种基于头脑风暴智能优化算法的BP神经网络模糊图像复原方法(OBSO-BP)。该方法在聚类和变异两方面优化了头脑风暴智能算法,利用头脑风暴优化算法易于解决多峰高维函数问题的特点,自动搜寻BP神经网络更佳的初始权值和阈值,以减少BP网络对其初始权值和阈值的敏感性,避免网络陷入局部最优解,增加网络的收敛速度,减小网络误差,提高图像还原质量。该文采用20张不同的图像,对其模糊图像分别进行维纳滤波复原(Wiener)、基于头脑风暴算法的维纳滤波复原(Wiener-BSO)、BP神经网络复原以及基于头脑风暴算法的BP神经网络(BSO-BP)图像复原实验。实验结果表明,该方法能够取得更好的图像复原效果。A kind of restoration method of BP neural network fuzzy image based on Optimized Brain Storming intelligent Optimized(OBSO-BP)algorithm is proposed in this paper.With the method of brain storming intelligent optimized algorithm which is optimized in both clustering and variation,issues of multi-peak highdimensional function is easily solved.This method optimizes brain storming intelligence algorithm from two aspects of clustering and mutation.This method makes use of the characteristics of brain storming optimization algorithm,which is easy to solve multi-peak and high-dimensional function problems,to automatically search for better initial weights and thresholds of BP neural network,thus reducing the sensitivity of BP network to its initial weights and thresholds,avoiding the network falling into local optimal solution,increasing the convergence speed of the network and reducing the network error and improving the quality of image restoration.Twenty different images are adopted to the image restoration experiment of their fuzzy images with Wiener filtering restoration(Wiener),Wiener filtering restoration based on optimized Brain Storming intelligent Optimized algorithm(Wiener-BSO),BP neural network restoration and BP neural network restoration based on optimized Brain Storming intelligent Optimized algorithm(BSO-BP).Results show that a better effect of image restoration can be achieved with this method.

关 键 词：图像复原 BP神经网络 头脑风暴算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]