基于可调滤波器金字塔算法的人脸特征提取

Face Feature Extraction Based on Steerable Filters

特征提取是人脸识别中最重要的一个环节,我们在HMAX模型的基础上,提出了一种改进的视觉信息处理模型,它利用可调滤波器金字塔算法对图像进行多尺度、多方向分析,并结合一系列分层处理,得到了具有尺度、平移、旋转不变性的特征向量,解决了人脸识别中的一个重要难题。在ORL人脸数据库中的实验结果证明了该算法的有效性。

Face feature extraction is the key procedure in face recognition. This paper presents an improved model of visual information processing, which analyses the image in different size and orientation using steerable filters. And a series of hierarchical processing are used. The feature vector can be extracted which is invariant in size, translation and rotation. It solves a difficulty in face recognition. Finally, the experiments on ORL face database verify the effectiveness of the proposed method.