基于2D-Gabor与KLDA的特征提取

提出一种2D-Gabor小波与核线性鉴别分析(KLDA)相结合的特征提取方法。该方法对经过预处理的人脸图像进行多方向、多尺度的2D-Gabor滤波，将滤波后的图像看作独立样本加入原样本库中，对新样本利用KLDA方法进行二次特征提取，得到较理想的类内聚度和类间散度样本特征，再采用三阶近邻分类器进行特征分类处理。实验结果表明，该方法相比传统方法识别率更高，易于工程实现。

Feature Extraction Based on 2D-Gabor and Kernel Linear Discriminant Analysis

This paper proposes a feature extraction method combining 2D-Gabor wavelet and Kernel Linear Discriminant Analysis（KLDA）. The pretreated face images are filtered with multi-scale and multi-orientation, and the filtered images are added into the original face database as separate samples to increase the number of samples. The classical KLDA method is applied to extract features once more to obtain the ideal sample characteristics of class cohesion and between-class scatter. Third-order nearest neighbor classifier is used to classify the features. Experimental results indicate that the method can get a better performance and recognition rate, and it is easy to implement in projects.