压缩感知中基于变尺度法的贪婪重构算法的研究

首先将最优化方法中的梯度法与贪婪算法相结合，在已有梯度追踪算法的基础上提出了基于变尺度法的梯度追踪（VMMGP）算法，分析比较这些算法的计算复杂度和存储需求；其次用这些算法分别重构一维、二维信号，分析比较了算法重构效果及其性能好坏。实验结果表明，对于一维信号和二维信号的重构，所提出的VMMGP算法，重构效果是所有梯度追踪算法中最好的。

Research on greedy reconstruction algorithms of compressed sensing based on variable metric method

A variable metric method based gradient pursuit algorithm based on some existing gradient pursuit algorithms was proposed by combining the gradient direction of optimization methods with greedy algorithms. The computational complexity and storage requirements of these algorithms were also analysed and compared. In experimental part, all these algorithms were used to reconstruct one-dimensional audio signal and two-dimensional image signal. Experimental results show that the VMMGP algorithm has the best reconstruction performance among all the gradient pursuit algorithms.