摘要
针对传统图像去运动模糊方法易放大噪声,产生振铃效应等问题,提出一种基于梯度域和深度学习的图像运动模糊盲去除算法。该算法利用引导滤波和L0滤波对图像进行预处理,将预处理后的梯度域图像块送入设计的卷积神经网络进行训练;提取训练好的模型参数,实现模糊核估计与图像复原;在图像复原过程中使用TV正则项进行图像去模糊。与其他算法相比,该算法能有效地抑制振铃效应和减弱噪声,去运动模糊效果较好。
Aiming at the problem of traditional image motion deblur which may amplify noise and produce ringing effect,a new image motion deblurring algorithm based on gradient domain and deep learning is proposed. In this proposed algorithm,the image is preprocessed by guided filtering and L0 filtering,and the preprocessed gradient domain image block is sent to the designed convolutional neural network for training. Parameters in the trained model are extracted to achieve fuzzy kernel estimation and image restoration. In the image restoration process,TV items are used to deblur the image. By comparing with the other algorithms,this proposed algorithm can effectively suppress the ringing effect and reduce the noise,and the motion blur effect is better.
作者
郭业才
郑慧颖
叶飞
GUO Yecai;ZHENG Huiying;YE Fei(Jiangsu Key Laboratory of Meteorological Observation and Information Processing,Nanjing University of Information Science & Technology,Nanjing 210044,China;College of Electronical and Information Engineering,Nanjing University of Information Science & Technology,Nanjing 210044,China)
出处
《实验室研究与探索》
CAS
北大核心
2019年第6期4-8,共5页
Research and Exploration In Laboratory
基金
国家自然科学基金项目(61673222)
江苏省高等教育教学改革研究课题(2017JSJG168)
关键词
运动模糊
卷积神经网络
梯度域
图像复原
motion blur
convolution neural network
gradient domain
image restoration
