结合小波分析和LBP算子的人脸描述与识别

Human Face Description and Recognition Combining Wavelet Analysis and LBP Operator

为提高人脸识别的准确性和实用性,提出了一种结合小波分析和LBP算子的人脸描述与识别算法.先利用小波分析对原始人脸图像进行降维,再分块求取小波系数的2类LBP直方图,最后将所有区域的2类LBP直方图连接起来得到整幅图像的小波直方图序列特征(HSWLBP),并将其作为人脸的鉴别特征用于分类识别.所提出的算法在ORL人脸数据库上取得高达0.99的人脸识别率.实验分析表明,HSWLBP具有较强的特征表示能力和可鉴别性,且对光照、人脸表情和位置的变化具有较高的鲁棒性.

In order to promote the accuracy and practicability of face recognition, a human face description and recognition method combining wavelet analysis and LBP operator is proposed in this paper. Firstly, wavelet analysis is employed to decompose the original face image for dimension reduction. Then the approximate image is divided into several regions from which two classes LBP feature distributions are extracted and respectively concatenated into two histogram sequences. Finally, the two sequences are concatenated into one enhanced vector which is the proposed histogram sequence of wavelet local binary patterns （HSWLBP） and is used as the face descriptor for classification and recognition. Experimental results on ORL face database show that the proposed method can achieve high face recognition rate which is up to 99%, and with high computational speed. This work demonstrates that HSW LBP feature is highly discriminable with good performance in feature expression and is robust to illumination, face expression and position variations.