基于图像矩阵判别局部保持投影的人脸识别

Face recognition based on image matrix discriminant locality preserving projections

提出一种基于图像矩阵判别局部保持投影的人脸识别方法。图像矩阵判别局部保持投影是在局部保持投影基础上进行了扩展,考虑了类标签信息并在其目标函数中增加类间散度约束,使得求解的特征更具判别性。另外,图像矩阵判别局部保持投影是直接处理图像矩阵而不需要将矩阵转化为向量,保留了像素间的空间位置关系,避免了奇异性问题。实验结果表明该方法是有效的。

A new face recognition method based on Image Matrix Discriminant Locality Preserving Projections（IMDLPP） is proposed.Based on Locality Preserving Projections（LPP）,IMDLPP takes into account the class label information,and adds betweenclass scatter constraint into the objective function.In addition,IMDLPP works directly with original image matrices which do not need to be converted into vectors,keeps the spatial position information of pixel in face image,and avoids the singular problem.The experimental results indicate the effectiveness of the proposed method.