Abstract: The traditional pre-processing prior to feature extraction method can not be used for partial face image processing，can’t analyze a region of interest and subject to background environmental impact，this paper presents a face image adaptive pre-processing method.This method identifies the human eye position from the human face image by two-dimensional Gabor filter，and then extractes region of interest by image segmentation，zooms the image，uses Principal Component Analysis（PCA） method to complete feature extraction，and finally the classification is done by the smallest neighbor classification two-dimensional，so the process of face recognition is completed.Experimental results show that face recognition method based on adaptive pre-processing can remove the hair，neck，shoulder and face which has nothing to do with the part of the face.It also increases the rate of face recognition，and has good robustness on the translation，rotation，scale changes and expression.

Key words: face recognition, adaptive pre-processing, Principal Component Analysis（PCA）, feature extraction, Gabor filter

摘要： 针对传统预处理方法在特征提取之前不能对人脸图像进行局部化处理，不能分析出感兴趣区域及受背景环境影响等缺点，提出一种人脸图像的自适应预处理方法。该方法通过二维Gabor滤波器从人脸图像中确定人眼位置，通过图像分割算法提取出感兴趣区域，缩放图像，运用主分量分析方法进行特征提取，通过二维最小近邻分类法进行分类，从而完成人脸识别过程。实验结果表明，基于自适应预处理的人脸识别方法能够有效去除头发、脖子、肩及与人脸无关的部分，提高了人脸识别率，且对一定的平移、旋转、尺度变化和表情有良好的鲁棒性。

关键词: 人脸识别, 自适应预处理, 主分量分析, 特征提取, Gabor滤波器