计算机应用研究
計算機應用研究
계산궤응용연구
APPLICATION RESEARCH OF COMPUTERS
2013年
3期
926-929
,共4页
人脸识别%镜像对称性%二维主成分分析%非迭代双边二维主成分分析%对称非迭代双边二维主成分分析
人臉識彆%鏡像對稱性%二維主成分分析%非迭代雙邊二維主成分分析%對稱非迭代雙邊二維主成分分析
인검식별%경상대칭성%이유주성분분석%비질대쌍변이유주성분분석%대칭비질대쌍변이유주성분분석
face recognition%mirror symmetry%two dimensional principal component analysis（2DPCA）%non-iteration bilate-ral projection based 2DPCA（NIB2DPCA）%symmetrical non-iteration bilateral projection based 2DPCA（SNIB2DPCA）
结合人脸图像的对称性在非迭代双边二维主成分分析（NIB2DPCA）的基础上, 提出了对称非迭代双边二维主成分分析（SNIB2DPCA）的人脸识别方法。该方法引入镜像变换, 根据奇偶分解原理分别生成奇、偶对称样本, 用NIB2DPCA分别对奇偶对称样本提取特征, 通过奇偶加权因子对奇偶对称样本的特征矩阵进行组合得到最终的分类特征矩阵, 最后用最近邻分类器分类。在Yale、ORL和YaleB人脸库上的实验表明该方法不仅显著提高了识别率, 而且对光照影响有一定的鲁棒性。
結閤人臉圖像的對稱性在非迭代雙邊二維主成分分析（NIB2DPCA）的基礎上, 提齣瞭對稱非迭代雙邊二維主成分分析（SNIB2DPCA）的人臉識彆方法。該方法引入鏡像變換, 根據奇偶分解原理分彆生成奇、偶對稱樣本, 用NIB2DPCA分彆對奇偶對稱樣本提取特徵, 通過奇偶加權因子對奇偶對稱樣本的特徵矩陣進行組閤得到最終的分類特徵矩陣, 最後用最近鄰分類器分類。在Yale、ORL和YaleB人臉庫上的實驗錶明該方法不僅顯著提高瞭識彆率, 而且對光照影響有一定的魯棒性。
결합인검도상적대칭성재비질대쌍변이유주성분분석（NIB2DPCA）적기출상, 제출료대칭비질대쌍변이유주성분분석（SNIB2DPCA）적인검식별방법。해방법인입경상변환, 근거기우분해원리분별생성기、우대칭양본, 용NIB2DPCA분별대기우대칭양본제취특정, 통과기우가권인자대기우대칭양본적특정구진진행조합득도최종적분류특정구진, 최후용최근린분류기분류。재Yale、ORL화YaleB인검고상적실험표명해방법불부현저제고료식별솔, 이차대광조영향유일정적로봉성。
This paper proposed a new algorithm called SNIB2DPCA, which combined the theory of NIB2DPCA with frontal facial symmetry. Firstly, it introduced mirror transform. After that, it decomposed original face samples into even symmetrical images and odd symmetrical ones through the theory of odd/even decomposition. Then employed NIB2DPCA to extract feature information from odd and even symmetrical samples separately. After that, combined the odd and even feature matric to form the final feature matric by the odd/even weighted factors. Finally, it employed the nearest neighbor classifier to classify the final feature matric. The method was evaluated on the Yale , ORL and YaleB face image databases. Both theoretical analysis and experimental results demonstrate that the proposed method not only significantly raises the recognition rate, but also has certain robustness to the influence of light.
