Zhang Qian, Li Hai Gang, Li Ming, Ding Lei
School of Information and Control Engineering, China University of Mining and Technology, Xuzhou 221116, China.
Math Biosci Eng. 2019 Dec 5;17(2):1578-1592. doi: 10.3934/mbe.2020082.
Affected by illumination, gesture, expression and other factor's variation, face image pattern is easy to be changed, so it is important to find a robust data representation for the correct classification of face pattern. In this paper, a face image recognition algorithm based on 2-D Gabor wavelet transform and Local Binary Pattern (LBP) is proposed. LBP is a local describe operator, which is invariant against illumination variation. 2-D Gabor wavelet transform have the invariant property against pose and expression variation. Experimental results show that the large scale 2-D Gabor wavelet representation could get good classification accuracy. Using LBP to describe 2-D Gabor wavelet representation of face image, together with image block, histogram statistics, PCA dimensionality reduction, nearestneighbors classification, we finally find this algorithm can get a better classification performance in different scales and directions.
受光照、姿态、表情等因素变化的影响，人脸图像模式容易发生改变，因此找到一种鲁棒的数据表示方法对于正确分类人脸模式至关重要。本文提出了一种基于二维伽柏小波变换和局部二值模式（LBP）的人脸图像识别算法。LBP是一种局部描述算子，它对光照变化具有不变性。二维伽柏小波变换对姿态和表情变化具有不变性。实验结果表明，大规模二维伽柏小波表示能够获得良好的分类准确率。利用LBP描述人脸图像的二维伽柏小波表示，结合图像分块、直方图统计、主成分分析降维、最近邻分类，最终发现该算法在不同尺度和方向上能够获得更好的分类性能。
