电子学报 ›› 2017, Vol. 45 ›› Issue (5): 1226-1233.DOI: 10.3969/j.issn.0372-2112.2017.05.028

• 学术论文 • 上一篇    下一篇

基于模糊结构图的模糊核估计

方帅1,3, 刘远东1, 曹洋2, 刘永进3   

  1. 1. 合肥工业大学计算机与信息学院, 安徽合肥 230009;
    2. 中国科学技术大学自动化系, 安徽合肥 230026;
    3. 光电控制技术重点实验室, 河南洛阳 471009
  • 收稿日期:2016-02-25 修回日期:2016-05-05 出版日期:2017-05-25
    • 作者简介:
    • 方帅 女,1978年1月出生于安徽寿县.博士,教授,合肥工业大学硕士生导师,主要研究方向为计算机视觉、图像复原等.E-mail:fangshuai@hfut.edu.cn
    • 基金资助:
    • 国家自然科学基金 (No.61175051,No.61175033); 安徽省自然科学基金 (No.1508085SMF222); 光电控制技术重点实验室和航空科学基金联合资助项目 (No.201451P4007)

Blur Kernel Estimation Using Blurry Structure

FANG Shuai1,3, LIU Yuan-dong1, CAO Yang2, LIU Yong-jin3   

  1. 1. School of Computer and Information, Hefei University of Technology, Hefei, Anhui 230009, China;
    2. Department of Automation, University of Science and Technology of China. Hefei, Anhui 230026, China;
    3. Key Laboratory of Optoelectronic Control Technology, Luoyang, Henan 471009, China
  • Received:2016-02-25 Revised:2016-05-05 Online:2017-05-25 Published:2017-05-25
    • Supported by:
    • National Natural Science Foundation of China (No.61175051, No.61175033); Natural Science Foundation of Anhui Province (No.1508085SMF222); Project supported by the Key Laboratory of Optoelectronic Control Technology and Aviation Science Foundation (No.201451P4007)

摘要:

图像结构边缘对模糊核估计有重要意义.近年来许多成功的算法都致力从潜在清晰图像中分离出结构边缘形成中间图像,然后用其与模糊图像一起估计模糊核.但是这些算法忽视了从模糊图像中分离出结构边缘对应的部分,导致核估计过程中目标函数的数据项不平衡.针对这一问题,本文利用中间图像和潜在模糊核产生二值模板对模糊图像进行处理,分离出结构边缘对应的部分,并用其修正目标函数.此外本文提出采用L0范数同时约束幅值域和梯度域的正则项,从而缩小核估计的解空间.多个标准测试数据库上实验结果表面,本文算法无论在鲁棒性还是准确性方面均具有更好的效果.

关键词: 去模糊, 反卷积, 模糊核, 图像复原

Abstract:

It has been proven that structure of image play an important role in kernel estimation.In recent years,many successful algorithms propose to generate intermediate image by extracting structure from latent image,and then use it for blur kernel estimation.However,these methods ignore to extract the correspondence from input blurry image.This will cause unbalanced data item of objective function.In this paper we first exploit a mask determined by convolution of intermediate image with kernel to generate the correspondence,and then take it into data item instead of blurry image to overcome the problem.Moreover,we have found that kernel shows the properties of sparse both in intensity domain and derivatives domain.Accordingly,we apply L0-norm regularization to constrain both intensity domain and derivatives domain of kernel.Compared with the state-of-the-art algorithms,experiments across datasets showed that our algorithm achieved better performance.

Key words: deblurring, deconvolution, blur kernel, image recovery

中图分类号: