Abstract: By introducing between-class scatter constraint and label information into two-dimensional locality preserving projections（2D-LPP） algorithm, two-Dimensional Discriminant Locality Preserving Projections（2D-DLPP） has more discriminant power than 2D-LPP. However, 2D-DLPP is confronted with the difficulty of parameter selection, which limits its power on solving recognition problem. To solve this problem, by constructing parameter-less affinity matrix, an algorithm called parameter-less two-dimensional discriminant locality preserving projections（parameter-less 2D-DLPP） is proposed. The simulation results on Yale and ORL face database show that the method in this paper can get higher recognition rate than 2D-DLPP, 2D-LPP and 2D-LDA.

Key words: face recognition, feature extraction, two-dimensional locality preserving projections, parameter-less

摘要： 通过向二维局部保持投影（2D-LPP）算法中引入类间约束和类标识信息，得到二维判别局部保持投影（2D-DLPP）算法，使它拥有更多的判别信息。但它却面临复杂的参数选择问题，这使得它在解决识别问题时受到限制。为解决此问题，构造无参数的相似矩阵，提出无参数的二维判别局部投影（无参数2D-DLPP）算法。在Yale和ORL人脸库上的仿真实验结果表明，该算法与二维判别局部保持投影（2D-DLPP）、二维局部保持投影法（2D-LPP）和二维线性判别分析法（2D-LDA）相比能够取得更高的识别率。

关键词: 人脸识别, 特征提取, 二维判别局部保持投影, 无参数