Computer Engineering and Applications ›› 2007, Vol. 43 ›› Issue (22): 25-27.

• 博士论坛 • Previous Articles     Next Articles

Method of image reconstruction based on very sparse random projection

FANG Hong1,2,ZHANG Quan-bing1,WEI Sui1   

  1. 1.Key Lab of Intelligent Computing and Signal Processing,Anhui University,Hefei 230039,China
    2.College of Science,Hefei University of Technology,Hefei 230009,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2007-08-01 Published:2007-08-01
  • Contact: FANG Hong

基于非常稀疏随机投影的图像重建方法

方 红1,2,章权兵1,韦 穗1   

  1. 1.安徽大学 计算智能与信号处理教育部重点实验室,合肥 230009
    2.合肥工业大学 理学院,合肥 230009
  • 通讯作者: 方 红

Abstract: Introduces the very sparse random projection into Compressed Sensing(CS) theory and presents a new kind of CS measurement matrix:very sparse projection matrix.By the asymptotic normality for very sparse random projection distribution,proves that the new matrix satisfies the necessary condition for CS measurement matrix.Owing to its sparsity of structure,new matrix greatly simplifies the projection operation during images reconstruction,which greatly improving the speed of reconstruction.The results of simulated and real experiments show that under a certain condition,new matrix can acquire exact reconstruction.Last,the compare of reconstruction results respectively adopting new matrix,Gaussian measurement matrix and Bernoulli measurement matrix is conducted.

Key words: random projection, compressed, sparsity

摘要: 将非常稀疏随机投影引入可压缩传感CS(Compressed Sensing)理论,提出一种新的CS测量矩阵:非常稀疏投影矩阵。利用非常稀疏投影分布的渐近正态性,证明了新的矩阵满足CS测量矩阵的必要条件。该矩阵由于其构成的非常稀疏性大大简化了图像重建过程中的投影计算,从而提高重建速度。实验结果表明非常稀疏投影矩阵在满足一定测量数目要求的条件下可以精确重建。最后给出了新的测量矩阵与一般采用的高斯和贝努里测量矩阵的重建结果比较和分析。

关键词: 随机投影, 可压缩, 稀疏性