摘要
提出了一种基于 BP 神经网络的图像复原算法。在分析图像模糊机制的基础上,为了降低输入维数,该方法采用滑动窗口操作来提取特征,同时为了加快训练速度和改善网络复原效果,首先对图像进行边缘提取,对图像内边缘区域和平坦区域分别采用滑动窗口获得训练集.利用 BP 神经网络的学习能力,通过训练,建立含有退化信息(高斯模糊)的模糊图像和清晰图像之间的映射关系模型,利用该模型对模糊图像进行复原,得到的复原图像在视觉上和定量分析上都获得了较好的效果.
Based on BP neural network,an image restoration method is presented.To reduce the input dimension,a sliding window method is employed to obtain the features of the blurred image.For the purpose of accelerating training and improving the restoration performance,the sliding window is applied to the edge part and smooth part to get the training sets,respectively.A mapping model between blurred image and clear image is established through training the BP neural network and then it is used to restore the blurred image.At the end, simulation experiments are performed and the results show that this model is practical.
出处
《红外与激光工程》
EI
CSCD
北大核心
2006年第z4期121-125,共5页
Infrared and Laser Engineering
基金
上海市教委基金(05NZ20)
中国博士后科学基金(2005037503)
关键词
图像复原
BP神经网络
滑动窗口
特征提取
Image Restoration
BP neural network
Sliding window
Feature extraction
