Abstract: Current face recognition algorithms lack robustness.In this paper，a novel feature extraction approach is firstly proposed to extract scale and pose invariant facial features，and then we have integrated this feature extraction approach with Laplacianface face recognition algorithm and built an improved Laplacianface face recognition algorithm based on shape-color features.Experimental results indicate that the proposed algorithm not only improves the accuracy of current face recognition algorithm，but also maintains higher recognition rate with variant face pose.Thus，the robustness of face recognition algorithm is also increased.

Key words: facial feature, AAM（Active Appearance Model）, color histogram, LPP（Locality Preserving Projection）, laplacianface,
face recognition

摘要： 针对人脸识别算法中普遍存在的鲁棒性不高的问题，提出一种新颖的特征提取手段，使提取的特征相对于图像尺度，人脸姿态等条件具有不变性;同时，将特征提取算法集成至Laplacianface人脸识别算法中，形成一种改进的基于形状-色彩特征的人脸识别算法。实验结果表明算法不仅提高了现有人脸识别算法的准确度，而且在人脸姿态等条件发生变化的情况下仍然能保持较高的识别率，有效提高了算法的鲁棒性。

关键词: 人脸特征, AAM（主动外观模型）, 颜色直方图, LPP（局部保持映射）, Laplacianface, 人脸识别