J4 ›› 2012, Vol. 39 ›› Issue (1): 105-110.doi: 10.3969/j.issn.1001-2400.2012.01.019

• 研究论文 • 上一篇    下一篇

具有随机位移的多帧图像超分辨重建快速算法

宁贝佳;冀峰;高新波   

  1. (西安电子科技大学 智能感知与图像理解教育部重点实验室,陕西 西安  710071)
  • 收稿日期:2011-09-11 出版日期:2012-02-20 发布日期:2012-04-06
  • 通讯作者: 宁贝佳
  • 作者简介:宁贝佳(1971-),男,讲师,西安电子科技大学博士研究生,E-mail: bjning@xidian.edu.cn.
  • 基金资助:

    陕西省科技攻关计划资助项目(2010K06-12);陕西省自然科学基础研究计划资助项目(2009JM8004)

Fast super-resolution reconstruction algorithms for multi-frame images with random shifts

NING Beijia;JI Feng;GAO Xinbo   

  1. (Ministry of Education Key Lab. of Intelligent Perception and Image Understanding,  Xidian Univ., Xi'an  710071, China)
  • Received:2011-09-11 Online:2012-02-20 Published:2012-04-06
  • Contact: NING Beijia

摘要:

常用的多帧图像超分辨重建算法大都基于图像恢复原理,因此计算量大,难以在线实时实现.为此,针对稀疏采样以及具有随机位移情况下的多帧图像超分辨重建问题,提出了三种融合重建快速算法,即邻域组合插值(NCI)方法、邻域扩展插值(NEI)方法以及全局双向线性插值(GTDLI)方法.仿真实验结果表明了三种算法的有效性,而且相对于经典的图像超分辨重建算法具有明显的速度优势.最后通过分析分别给出了三种算法所适合的应用场合.

关键词: 超分辨重建, 图像融合, 随机位移

Abstract:

The traditional multi-frame super-resolution (SR) reconstruction methods are computationally expensive and only applicable to offline processing cases. To reduce the workload of SR fusion for real-time applications, three algorithms, i.e., neighborhood combination interpolation (NCI), neighborhood expansion interpolation (NEI) and global two-direction linear interpolation (GTDLI), are proposed in this paper under the condition of multiple sparsely-sampled lower resolution images with global random shifts. Simulation results illustrate the effectiveness of these proposed algorithms and the advantages in speed over classic SR algorithms. Finally, the applicable fields for the three algorithms are concluded through theoretical analysis and experimental comparison.

Key words: super-resolution reconstruction, image fusion, random shifts