东北大学学报(自然科学版) ›› 2021, Vol. 42 ›› Issue (4): 531-537.DOI: 10.12068/j.issn.1005-3026.2021.04.011

• 信息与控制 • 上一篇    下一篇

基于模拟退火粒子群算法的MIT图像重建方法

杨丹1,2,3, 芦甜1,2, 郭文欣1, 王旭1   

  1. (1. 东北大学 信息科学与工程学院, 辽宁 沈阳110819; 2. 东北大学 辽宁省红外光电材料及微纳器件重点实验室, 辽宁 沈阳110819; 3. 东北大学 智能工业数据解析与优化教育部重点实验室, 辽宁 沈阳110819)
  • 修回日期:2020-09-11 接受日期:2020-09-11 发布日期:2021-04-15
  • 通讯作者: 杨丹
  • 作者简介:杨丹(1979-),女,辽宁营口人,东北大学副教授.
  • 基金资助:
    国家自然科学基金资助项目(51607029,61836011); 中央高校基本科研业务费专项资金资助项目(2020GFZD008,2020GFYD011).

MIT Image Reconstruction Method Based on Simulated Annealing Particle Swarm Algorithm

YANG Dan1,2,3, LU Tian1,2, GUO Wen-xin1, WANG Xu1   

  1. 1. School of Information Science & Engineering, Northeastern University, Shenyang 110819, China; 2. Key Laboratory of Infrared Optoelectric Materials and Micro-Nano Devices, Northeastern University, Shenyang 110819, China; 3. Key Laboratory of Ministry of Education on Data Analytics and Optimization for Smart Industry, Northeastern University, Shenyang 110819, China.
  • Revised:2020-09-11 Accepted:2020-09-11 Published:2021-04-15
  • Contact: YANG Dan
  • About author:-
  • Supported by:
    -

摘要: 为了改善逆问题病态性又能提高图像重建质量,提出了一种基于模拟退火粒子群算法的MIT图像重建方法.根据Hessian矩阵的维度,构建了一种Tikhonov和NOSER型混合多参数正则化算法.将模拟退火算法和粒子群算法进行组合,以广义交叉准则构建目标函数,进行正则化多参数寻优.结果表明,所提方法不仅有效克服了MIT重建图像数值解的不稳定性,增强了抗噪性能,而且所获得的重建图像的质量优于Tikhonov正则化和混合正则化算法,为MIT技术应用提供了理论参考.

关键词: 逆问题病态性;图像重建;Hessian矩阵;模拟退火;粒子群算法

Abstract: In order to improve the ill-posed inverse problem and improve the quality of image reconstruction, a MIT image reconstruction method based on simulated annealing and particle swarm optimization was proposed. According to the dimensions of the Hessian matrix, a Tikhonov and NOSER hybrid multi-parameter regularization algorithm was constructed. The simulated annealing algorithm and particle swarm algorithm were combined, the objective function was constructed by the generalized cross criterion, and the regularized multi-parameter optimization was performed.The results show that not only the proposed method effectively overcomes the instability of the numerical solution of the MIT reconstructed image and enhances the anti-noise performance, but also the quality of the obtained reconstructed image is better than that of Tikhonov regularization and hybrid regularization algorithms, which provides a theoretical reference for the application of MIT technology.

Key words: ill-posed inverse problem; image reconstruction; Hessian matrix; simulated annealing; particle swarm optimization

中图分类号: