基于自适应重启方法的快速压缩感知算法.pdfVIP

第 12 卷第 4 期 .:L f{ 弛3制b究寿-l Yo l. 12,No. 4 2015 年 7 月 July , 2015 文章编号 :1672一7940(2015 )04一0556-05 doi:10 . 3969/ j . issn.1672 一7940.2015.04. 025 基于自适应重启方法的快速压缩感知算法 王帅， 张猛，张俊宇 (华中师范大学计算机学院，湖北武汉 430079) 摘 要: NESTA 是一种解决压缩感知问题的快速准确一阶优化算法.提出了一种自适应量启方法来提高 NESTA算法的收敛速度. 新方法自动检测目标函数在优化过程收敛速度减缓的趋势，重量算法的参数，并 使算法从当前步骤重新开始运行. 实验结果表明该方法能够显著提离经典算法的收敛速度，从而提升算法运 行效率. 关键词: 压缩感知;NESTA;梯度下降;图像重构;重启算法;收敛速度 中固分类号: TP391 文献标识码: A 收稿日期: 2015 一03-02 The Fast Compressed Sensing Algorithm ßased on Adaptive Restart Method W a ng Shuai , Zhang M eng , Zha ng Junyu (School 01 Comþuter Science , Central China Normal University , Wuhan Hubei 430079 , China) Abslracl: NESTA is a fast and accur ate first-order optimization algorithm to solve problem of compressed sensing. T h e paper proposes an adaptive restart method to improve the con- vergence rate of NESTA. As object function decreases slowly in NESTA , this method can cope with it and significantl y improve the convergence rate of the algorithm. Making the al- gorithm restart from current step , the adaptive restart strategy automatically detects the slow convergence rate of object function and r esets the par ameters of the algorithm. Exper- imental results show that the method can significantly improve the convergence r ate of the classical algorithm , thereby enhance the efficiency of th e algorithm. Key words: compressed sensing; NESTA; gradient d escent; image reconstruction; restart method; convergence rate