«上一篇/Previous Article|本期目录/Table of Contents|下一篇/Next Article»

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:

期数:

2013年08期

页码:

234-237

栏目:

应用开发研究

出版日期:

1900-01-01

文章信息/Info

Title:

Improved Matching Pursuit Algorithms and Application in Image Compression

文章编号:

1673-629X（2013）08-0234-04

作者:

陆望; 王友国

南京邮电大学 理学院

Author(s):

LU Wang; WANG You-guo

关键词:

压缩传感; 匹配追踪; 重建算法; 图像压缩

Keywords:

compressive sensing; matching pursuit; reconstruction algorithm; image compression

文献标志码:

A

摘要:

压缩传感，是近年来新出现的一种采样定理。它的特点是对信号进行采样所需要的条件远远小于Nyquist采样速率。这种采样定理要求信号是稀疏的或者是可压缩的，并能在采样时对信号数据进行压缩，然后通过非线性重建算法完美重建信号。它突破了Nyquist采样定理，因此具有广阔的发展前景。重建算法中有一类称为匹配追踪算法，文中围绕改进的匹配追踪算法在图像压缩中的应用展开了研究，对OMP算法、ROMP算法进行了实现，并对算法本身以及其重构效果做出了比较；针对按列处理速度较慢的缺点，使用了分块处理的方法，降低运算时测量矩阵的规模，实验表明，分块处理确实能够加快运算速度。由于自然信号进行稀疏变换后，稀疏度不确定，造成重构时迭代次数不够合理。针对这个现象，文中提出了如何确定合适的迭代次数的方法，提高重建的精确度。这个方法本身会消耗时间，可以在权衡了重构精确度要求和时间要求后确定是否使用

Abstract:

Compressive Sensing ( CS) is a new signal sampling theory which breaks through the Nyquist sampling theory,and has a broad development prospect. The new method requires the signal to be sparse or compressible,and obtains the discrete samples of original sig-nal. It is able to compress a signal during the process of sampling,and then reconstructs the signal perfectly using the nonlinear reconstruc-tion algorithms. The Matching Pursuit ( MP) algorithm is one kind of the reconstruction algorithms. Research the applications of matching pursuit algorithm in the field of image compression,give the implementation of OMP,ROMP algorithm and the comparison of their re-construction results. To increase the convergence speed of OMP algorithm,and reduce matrix scale,the image is divided into blocks. The experience results show that the new scheme do improves the computing efficiency. At last,an idea of how to determine the proper itera-tive times is presented. Since the sparsity of a random signal is unknown,iterative times cannot be rational. The method itself is a iteration process,which consumes computing time. Use this method after measuring the requests of reconstruction accuracy and speed

常用功能

更新日期/Last Update: 1900-01-01