北京师范大学学报（自然科学版）2017，Vol.53Issue(1)：12-18,7.DOI:10.16360/j.cnki.jbnuns.2017.01.003
用于人脸识别的改进MKD-SRC方法
Face recognition via optimized MKD-SRC method
摘要
Abstract
Sparse representation is a hot topic in image processing,pattern recognition and computer vision.It has been widely applied in image compressing,image de-noising and restoration,target detection,object recognition,etc.For face recognition,a multi-task SRC method based on local features,the multikeypoint descriptors based SRC (MKD-SRC),is invariantly used to image rotating,scaling and translation,but it either involves high computational complexity or is not robust enough for illumination variations.Considering those problems,we examined the theory and premise of MKD-SRC,and propose an optimized MKD-SRC method based on filtering linear subspace and maximum likelihood probability.The proposed method has been estimated on public face databases.Experimental results showed its efficiency and robustness against large block occlusion and non-uniform illumination.关键词
人脸识别/稀疏表示分类方法/改进MKD-SRC/线性子空间/极大似然概率Key words
face recognition/sparse representation based classification method/optimized MKD-SRC method/linear subspace/maximum likelihood probability分类
数理科学引用本文复制引用
何珺,孙波..用于人脸识别的改进MKD-SRC方法[J].北京师范大学学报（自然科学版）,2017,53(1):12-18,7.基金项目
国家自然科学基金资助项目(61501035) （61501035）
中央高校基本科研业务费专项资金资助项目(2014KJJCA15) （2014KJJCA15）
