基于SURE无偏估计的图像自适应稀疏收缩去噪  

Image denoising with adaptive sparse shrinkage method based on SURE estimation

作　　者：沙正虎[1] 余剑[1] 崔琛[1,2] 

机构地区：[1]解放军电子工程学院信息工程系,合肥230037 [2]安徽省电子制约技术重点实验室,合肥230037

出　　处：《计算机工程与应用》2013年第7期195-198,212,共5页Computer Engineering and Applications

基　　金：国家自然科学基金(No.61040007)

摘　　要：研究了基于过完备字典下稀疏表示框架的图像去噪问题,基于Stein无偏估计提出一种自适应阈值选择算法。在一阶可导收缩函数的基础上,推导了阈值选择的优化目标函数;证明该函数是关于阈值的凸函数,用黄金分割法搜索其全局最小值。该算法选择的阈值接近峰值信噪比-阈值曲线的极大值点。通过对Lena和Barbara图进行去噪实验,验证了算法的优越性。Under sparse representation framework, image denoising problems are investigated over over-complete dictionary and a novel adaptive threshold selection method is presented based on Stein unbiased risk estimator. Based on the one order deriv- able shrinkage function, the optimal objective function about threshold selection is derived and it is shown to be convex function on threshold, and then its global minimum is searched by golden section method. The choice of threshold is closer to the peak signal to PSNR-threshold curve＇ s maxima. Lena and Barbara are used in experimental simulation, and the results demonstrate the superior performance of the algorithm.

关 键 词：稀疏表示 收缩去噪 通用阈值 SURE无偏估计 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TN911.73[自动化与计算机技术—计算机科学与技术]