基于调和模型的快速神经网络图像复原算法  

Fast Image Restoration Algorithm Using Neural Network Based on Harmonic Model

作　　者：吴亚东[1] 孙世新[2] 

机构地区：[1]西南科技大学计算机科学与技术学院,四川绵阳621010 [2]电子科技大学计算机科学与工程学院,四川成都610054

出　　处：《计算机应用研究》2007年第6期158-160,共3页Application Research of Computers

摘　　要：针对传统神经网络图像复原算法在复原过程中模糊图像边缘,收敛速度慢等不足,提出一种基于调和模型的快速神经网络图像复原算法。在该算法中,图像复原模型的正则化项采用调和模型,并在每次网络状态更新时引入最陡下降方法,使得网络能量迅速减小。实验表明,提出的算法能够很好复原图像的边缘特征,并具有快速收敛等优点。Traditional image restoration algorithm using neural network has some shortcomings was fond in experiments, such as blurring edges, low speed of convergence. A fast image restoration algorithm using neural network based on harmonic model was given. In this algorithm, this paper used harmonic model as regular term in image restoration model, and introduced the steepest descent method in update rule to accelerate the convergence of network. Experiment results show that the proposed algorithm could restore the degraded image while preserving the edge in a fast speed.

关 键 词：图像复原 神经网络 调和模型 去模糊 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]