改进的自适应权值核范数最小化去噪算法  

Improved self-adapting weighted nuclear norm minimization denoising algorithm

作　　者：刘玉兰 刘小平[1,2] 邹艳妮[1] 

机构地区：[1]南昌大学信息工程学院,江西南昌330031 [2]卡尔顿大学系统与计算机工程系

出　　处：《计算机工程与设计》2018年第1期212-217,229,共7页Computer Engineering and Design

基　　金：国家863高技术研究发展计划基金项目(2013AA013804)

摘　　要：为解决传统权值核范数最小化(WNNM)算法在最优参数选取过程中过度依赖经验值的问题,提出一种改进的自适应参数选取WNNM算法,其最大特点是在WNNM算法基础上增加了噪声评估模型。通过提取均值减损对比归一化系数和邻域系数的分布特征参数构成图像特征矢量,与其对应的噪声浓度共同组成样本集;利用支持向量回归对样本集进行训练得到噪声评估模型,快速有效地为算法提供最优参数。实验结果表明,相比传统WNNM算法,该算法在进行图像去噪时,效率更高,效果更好,具有良好的鲁棒性和泛化性。To solve the problem that the selection of the noise parameter relies heavily on experience value,an improved WNNM algorithm for adaptive parameter selection was proposed.The advantage of the algorithm was to build the noise evaluation model based on the traditional WNNM algorithm.The image features were constructed by combining the statistical features of the mean subtracted contrast normalized(MSCN) coefficients and its four direction neighboring MSCN coefficients,and the sample set was constituted by the image feature and the noise concentration corresponding to it.The sample set was trained using the support vector regression method to obtain the noise evaluation model,which provided the optimal parameter for WNNM algorithm.Experimental results show that the proposed algorithm is more efficient and has better robustness and generalization performances than traditional method.

关 键 词：%WNNM算法 参数选取 噪声评估模型 参数自适应 高斯噪声 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]