计算机技术与发展2017，Vol.27Issue(1)：190-194,5.DOI:10.3969/j.issn.1673-629X.2017.01.043
基于梯度下降法与QR分解的观测矩阵优化
Optimization of Measurement Matrix Based on Gradient Descent Method and QR Decomposition
摘要
Abstract
The design of the measurement matrix is the core of the theory of Compressive Sensing ( CS) . Based on the CS,the perform-ance of the measurement matrix is improved by the way which is that the correlation between the measurement matrix and sparse trans-formed matrix is reduced and that the column independence of the measured matrix is increased. Depending on this,a method to improve the performance of the observation matrix is proposed. The Gram matrix is processed by the gradient descent method to reduce non-diag-onal elements and the matrix obtained from the last step is dealt by the QR decomposition. The simulation experiment is carried out on the measurement matrix to test the validity of the algorithm. The result shows that the measurement matrix dealt by this method has better per-formance in Peak Signal-Noise Ratio (PSNR) and stability of reconstruction especially when the compression ratio is 0. 30,and the PSNR of this matrix is 67% higher than the matrix without any treatments.关键词
压缩感知/观测矩阵/梯度下降法/QR分解Key words
compressive sensing/measurement matrix/gradient descent method/QR decomposition分类
信息技术与安全科学引用本文复制引用
兰明然,王友国,郑丹青,翟其清..基于梯度下降法与QR分解的观测矩阵优化[J].计算机技术与发展,2017,27(1):190-194,5.基金项目
国家自然科学基金资助项目(61179027) （61179027）
