计算机工程

• 图形图像处理 • 上一篇    下一篇

基于联合非负字典学习的遥感图像超分辨重建

魏巍,吴孔平,郭来功,秦蒙   

  1. (安徽理工大学 电气与信息工程学院,安徽 淮南 232001)
  • 收稿日期:2015-05-19 出版日期:2016-08-15 发布日期:2016-08-15
  • 作者简介:魏巍(1981-),男,讲师、硕士,主研方向为数字图像处理、智能算法;吴孔平,副教授、博士;郭来功,副教授、博士研究生;秦蒙,硕士研究生。
  • 基金项目:
    安徽理工大学青年教师科学研究基金资助项目(QN201311);安徽理工大学大学生创新创业训练计划基金资助项目(AH201410361178)。

Super-resolution Reconstruction of Remote Sensing Images Based on Joint Nonnegative Dictionary Learning

WEI Wei,WU Kongping,GUO Laigong,QIN Meng   

  1. (School of Electrical and Information Engineering,Anhui University of Science and Technology,Huainan,Anhui 232001,China)
  • Received:2015-05-19 Online:2016-08-15 Published:2016-08-15

摘要: 针对图像的超分辨重建问题,提出一种基于联合非负字典学习的单幅图像超分辨重建算法,并将其用于遥感图像的超分辨重建。利用已有的高分辨图像,通过预处理得到高低分辨样本集。给出联合非负字典学习技术,并采用该技术对高低分辨样本集训练得到稀疏的高低分辨字典,使用高低分辨字典重建高分辨图像,并分析算法的计算复杂度。实验结果表明,与双三次插值法、联合字典训练算法、耦合字典训练算法相比,该算法在保证较好重建效果的同时需要较小的计算量。

关键词: 联合字典训练, 稀疏表示, 超分辨重建, 遥感图像, 计算复杂度

Abstract: Aiming at the super-resolution reconstruction of images,a super-resolution reconstruction algorithm for a single image based on joint nonnegative dictionary learning is proposed in this paper,and it is applied in the super-resolution reconstruction of remote sensing images.Using existing high-resolution images,high-resolution and low-resolution samples are obtained by preprocessing.Joint nonnegative dictionary training technology is proposed,and high-resolution dictionary and low-resolution dictionary are obtained by training high-resolution and low-resolution samples,respectively.Super-resolution remote sensing image is recovered by these dictionaries,and computation complexity is analyzed.Experimental results show that,compared with bicubic interpolation method,Joint Dictionary Training(JDT) algorithm and coupled dictionary training algorithm,the proposed algorithm requires lower computation cost to achieve better reconstruction effect.

Key words: Joint Dictionary Training(JDT), sparse representation, super-resolution reconstruction, remote sensing image, computation complexity

中图分类号: