东北大学学报:自然科学版 ›› 2017, Vol. 38 ›› Issue (4): 486-492.DOI: 10.12068/j.issn.1005-3026.2017.04.007

• 信息与控制 • 上一篇    下一篇

基于块分割的新型压缩感知算法

张娜1, 曹琨2, 刘亚轩1   

  1. (1. 中国海洋大学 信息科学与工程学院, 山东 青岛266100; 2. 河南牧业经济学院 信息与电子工程系, 河南 郑州450044)
  • 收稿日期:2015-10-27 修回日期:2015-10-27 出版日期:2017-04-15 发布日期:2017-04-11
  • 通讯作者: 张娜
  • 作者简介:张娜(1980-),女,山东青岛人,中国海洋大学讲师,博士.
  • 基金资助:
    山东省自然科学基金资助项目(ZR2014FQ027); 中央高校基本科研业务费专项资金资助项目(201513015).

New Compressive Sensing Algorithm Based on Block Segmentation

ZHANG Na1, CAO Kun2, LIU Ya-xuan1   

  1. 1. College of Information Science and Engineering, Ocean University of China, Qingdao 266100, China; 2. Department of Information and Electronic Engineering, Henan University of Animal Husbandry and Economy, Zhengzhou 450044, China.
  • Received:2015-10-27 Revised:2015-10-27 Online:2017-04-15 Published:2017-04-11
  • Contact: ZHANG Na
  • About author:-
  • Supported by:
    -

摘要: 针对现有块分割压缩感知(block compressive sensing,BCS)算法的块效应问题,提出一种低复杂度、可消除块效应的新型块分割重构算法.在稀疏表达时,采用小波变换(DWT)代替离散余弦变换(DCT),改善图像细节分量;在测量时,依据分块图像频率特征对测量矩阵加权,提高图像质量;在重构时,采用正交匹配追踪(orthogonal matching pursuit,OMP)算法代替匹配追踪(matching pursuit,MP)算法,提高重构速度.仿真结果表明,所提出的算法可在保证重构速度的情况下,有效消除块效应,且不增加内存占用.

关键词: 压缩感知, 小波变换, 正交匹配追踪算法, 测量矩阵, 块效应

Abstract: To solve the blocking artifacts of prior block segment compressive sensing, a new reconstruction algorithm was proposed which could reduce the blocking artifacts at low complexity. When sparse representing, discrete wavelet transform (DWT) method was utilized instead of discrete cosine transform (DCT) to improve detail component of image. When measuring, the measurement matrix of each block was reweighted to improve the quality of image according to the difference frequency between each block. When reconstructing, the orthogonal matching pursuit (OMP) algorithm was used to speed up reconstruction rather than the matching pursuit (MP) algorithm. Simulation results demonstrated that the blocking artifacts could be effectively eliminated by the proposed algorithm without making any effects on reconstruction speed and memory requirement.

Key words: compressive sensing (CS), wavelet transform, orthogonal matching pursuit (OMP) algorithm, measuring matrix, blocking artifacts

中图分类号: