一种改进的非凸秩最小化算法及其在矩阵恢复中的应用

An Improved Non-convex Rank Minimization Algorithm and Its Application in Matrix Recovery

在分析现有处理矩阵恢复问题的非凸秩最小化算法的基础上,提出了一种基于超松弛迭代的改进算法,并给出了松弛因子ω的确定准则。仿真实验表明:在惩罚参数选取较大的情形下,改进算法较原算法具有更快的收敛速度及更高的收敛精度,同时展示了基于非凸秩最小化算法的矩阵恢复技术在图像去噪中的应用。

This paper proposes a kind of nonlinear successive over -relaxation improved algorithm to solve ma-trix recovery problems on the basis of analyzing the nature of existing non -convex rank minimization algo-rithm,and the modifying criteria of relaxation factor ωis also given.Experimental results show that the im-proved algorithm usually has much faster convergence rate and higher precision than the original one when the penalty parameter is relatively large .In addition,the paper expounds applications of matrix recovery technolo-gy based on non-convex rank minimization algorithms in image denoising .