TSCG算法在图像修复中的应用

Application of TSCG Algorithm in Image Restoration

霍等人提出了求解大规模无约束优化问题的三维子空间共轭梯度算法(简称为TSCG),该算法是在一个特殊三维子空间上,通过极小化目标函数的二次近似模型来确定嵌套参数。在温和的假设条件下,证明了该算法对一般非凸函数的全局收敛性,并通过应用在无约束优化问题的数值实验,表明了该算法具有鲁棒性和高效性。在此基础上,文中把该算法运用到添加三种不同噪声的图像修复问题中,表明了该算法在解决这类问题中的性能是稳健且高效的。

A three-dimensional subspace conjugate gradient algorithm(TSCG for short)is considered for solving large-scale unconstrained optimization problems by Huo et al.This algorithm determines the embedding parameters by minimizing the quadratic approximation model of the objective function in a special three-dimensional subspace.The global convergence of the proposed algorithm for general non-convex functions is proved under mild assumptions.Numerical experiments on unconstrained optimization problems show that the given algorithm is robust and efficient.In this paper,this algorithm is applied to the problem of image processing with three different noises,which shows that the performance of this algorithm is robust and efficient in solving this kind of problems.