人脸识别技术易受光照、表情等因素影响，为充分提取人脸特征信息，提出了融合改进的局部二值模式(LBP)和梯度方向直方图(HOG)方法提取人脸图形纹理、细节特征，利用列方向压缩的2DPCA+PCA算法对人脸的特征空间进行降维处理，使用2DPCA算法降低了特征维度，解决了仅仅使用PCA方法，由于人脸图像特征维度高而造成求解模型复杂的问题，降低了计算规模，提高了运算速度。最后，使用ORL和Yale人脸数据库进行实验。结果表明，基于改进的LBP和HOG融合的特征提取具有一定的互补性，与其它的识别算法相比，该改进的算法识别率有了较大的提高，鲁棒性更强。

Recognition rate of face recognition technology is vulnerable to illumination, expression and other factors. An image method based on the improved Local Binary Pattern (LBP) and Histogram of Oriented Gradient (HOG) for texture and details feature extraction was proposed, and 2DPCA+PCA algorithm of column direction compression was employed to reduce the dimensions of face feature space, so as to solve the problem that the PCA model was complex because of the high dimension of face image features, and to reduce the computing scale and furthermore to improve the computing speed. The testing results of experiments with ORL and Yale face database illustrate that the feature extraction based on the improved LBP and HOG fusion are complementary, and the recognition rate of the improved algorithm has been greatly improved compared with other recognition algorithms.