摘要
BP(Back-propagating)神经网络在图像处理中主要应用于模糊图像的复原,因为该算法在收敛速度上有一定的局限性,同时比较容易受局部极小值的影响.应用LMBP(LevenbergMarquardt BP)神经网络算法对模糊图像进行复原.这种算法实质上是提取了神经网络的GaussNewton法以及梯度下降法的优势,加速了算法收敛.实验结果表明,基于LMBP神经网络的图像复原方法对模糊图像的修复效果明显,且运行速度快.
BP(Back-propagating)neural network is mainly applied to the restoration of fuzzy images in image processing,because the algorithm has certain limitations in convergence speed and is more susceptible to local minima.To apply LMBP(Levenberg-The Marquardt BP)neural network algorithm recovers the blurred image.This algorithm essentially extracts the advantages of the Gauss-Newton method of the neural network and the gradient descent method,which accelerates the convergence of the algorithm.The experimental results show that the image restoration method based on LMBP neural network has obvious effect on the restoration of blurred images,and the algorithm runs fast.
引文
[1]赵建梁.基于卷积神经网络的图像复原方法研究[D].南京:南京大学,2018
[2]赵雪青.降质图像复原方法研究[D].西安:陕西师范大学,2013
[3]易丽娅.图像复原的Bregman迭代正则化方法研究[D].武汉:华中科技大学,2011
[4]Banham M R,Katsaggelos A K.Digital image restoration[J].IEEE signal processing magazine,1997,14(2):24-41
[5]jun Wang H,Tao J,Neimule M.Application of ACO-LMBP Hybrid Neural Network Algorithm in Image Denoising[C]//IOP Conference Series:Materials Science and Engineering.IOP Publishing,2018,439(3):032120
[6]赵旭峰.LMBP神经网络算法改进及其应用[D].合肥:中国科学技术大学,2014
[7]孙毅.基于BP神经网络的图像复原研究[D].兰州:甘肃政治学院,2018
[8]ZHANG Y,SUN L,YAN C,et al.Adaptive residual networks for high-quality image restoration[J].IEEE Transactions on Image Processing,2018,27(7):3150-3163
[9]YU K,DONG C,LIN L,et al.Crafting a toolchain for image restoration by deep reinforcement learning[C]//Proceedings of the IEEE conference on computer vision and pattern recognition.2018:2443-2452
[10]董俊.基于BP神经网络的图像复原算法研究[D].西安:西安科技大学
[11]李炯城,黄汉雄.神经网络中LMBP算法收敛速度改进的研究[J].计算机工程与应用,2006,42(16):46-49
[12]张妮,田学民.LMBP神经网络改进算法的研究[J].计算机与应用化学,2010,27(8):1111-1114