计算机应用与软件2017，Vol.34Issue(11)：204-209,216,7.DOI:10.3969/j.issn.1000-386x.2017.11.038
牛顿迫近迭代算法在图像恢复中的应用
APPLICATION OF NEWTON PROXIMAL ITERATIVE ALGORITHM FOR IMAGE RESTORATION
摘要
Abstract
The mixture object function is composed of smooth and non-smooth function.The traditional first-order optimization algorithm is limited by the first-order approximation of the smooth function and the search step.And it is difficult to obtain a high-precision solution of the objective function.Therefore,we propose a second order Newton proximal operator splitting iterative algorithm.Firstly,Taylor expansion of the smoothing function was used to obtain the two order transformation model of the objective function.The transformed model was decomposed into Newton iterative subproblem and proximal iterative subproblem.Then,the search direction and the optimization search step length of Newton iterative sub-problem were given.Finally,the convergence property was analyzed.Taking advantage of image blurred by system and noise for restoration,we perform the recovery experiments.The results show the PSNR (peak signal to noise ratio) of the proposed method is about 2 dB higher than other methods,and the SSIM (structural similarity index measure) is improved by about 3％.关键词
非光滑特性/牛顿迫近算法/迭代收敛/图像恢复Key words
Non-smooth property/Newton proximal algorithm/Iterative convergence/Image restoration分类
信息技术与安全科学引用本文复制引用
李旭超,刘燕,李玉叶..牛顿迫近迭代算法在图像恢复中的应用[J].计算机应用与软件,2017,34(11):204-209,216,7.基金项目
国家自然科学基金项目(11402039) （11402039）
2016年度内蒙古自治区科技厅自然科学基金项目(2016MS0602) （2016MS0602）
2016年度内蒙古自治区高等学校科学研究项目(NJZY16254). （NJZY16254）
