文献类型：会议论文

英文题名：Full Space and Subspace Methods for Large Scale Image Restoration

作者：Wang, Yanfei[1];Ma, Shiqian[2];Ma, Qinghua[3,4]

第一作者：Wang, Yanfei

通讯作者：Wang, YF[1]

机构：[1]Chinese Acad Sci, Inst Geol & Geophys, Beijing 100029, Peoples R China;[2]Columbia Univ, Dept Ind Engn & Operat Res, New York, NY 10027 USA;[3]Renmin Univ China, Sch Informat, Beijing 100872, Peoples R China;[4]Beijing Union Univ, Coll Arts & Sci, Beijing 100083, Peoples R China

第一机构：Chinese Acad Sci, Inst Geol & Geophys, Beijing 100029, Peoples R China

通讯机构：[1]corresponding author), Chinese Acad Sci, Inst Geol & Geophys, Beijing 100029, Peoples R China.

会议论文集：1st International Workshop on Optimization and Regularization for Computational Inverse Problems and Applications

会议日期：JUL 21-25, 2008

会议地点：Beijing, PEOPLES R CHINA

语种：英文

摘要：In this chapter, we discuss about the full space and subspace methods for ill-posed image restoration problems. Image restoration refers to minimizing the degradation which is caused by sensing environment, say CCD camera misfocus, nonuniform motion, atmospheric aerosols and atmospheric turbulence. For image restoration problems, a key matter is to solve a quadratic programming problem. We study numerical solution methods in full space by limited memory of BFGS method and the subspace trust region method. We develop a novel approach for reducing the cost of sparse matrix-vector multiplication when applying the full space and subspace methods to atmospheric image restoration. Also the projection technique for the regularized convex quadratic functional is developed in the iteration for ensuring nonnegativity. Numerical experiments indicate that these methods are useful for large-scale image restoration problems.