一种新的HOG特征人脸图像识别算法研究
|
摘要
针对人脸在非均匀光照下识别率的降低,提出了拉普拉斯滤波和离散余弦变换(DCT)融合梯度方向直方图(HOG)人脸识别算法。首先通过拉普拉斯滤波对人脸图像进行处理,突出其纹理特征;其次进行离散余弦变换(DCT),有效滤除高频分量;然后利用离散余弦逆变换(IDCT)重建人脸图像,降低其维数;最后通过梯度方向直方图(HOG)算子提取人脸图像固有特征,并利用最近邻方法进行分类识别。实验结果表明,该算法在不同特征维数下的Yale B人脸数据库中识别率高达95%以及课题组自建的维吾尔族人脸数据库中其识别率高达98.5%,优于其他传统算法,具有很强的鲁棒性和实时性。
For the reduction of human face recognition rate under non-uniform illumination,Laplace filter and discrete cosine transform (DCT) gradient direction histogram (HOG) face recognition algorithm is proposed. First by Laplace filtering of face image processing,its texture feature is highlighted; Secondly in discrete cosine transform (DCT),high frequency components are effectively filtered out; Then using discrete cosine inverse transformation (IDCT) reconstruction of face image reduces its dimension; Finally through the gradient direction histogram (HOG) operator the inherent characteristics of facial image recognition are extracted and the nearest neighbor method is used. The experimental results show that the proposed algorithm under different characteristic dimension of human face recognition rate as high as 95% in the database and Yale B group of uighurs face database construction in its recognition rate is as high as 98.5%,is better than other traditional algorithm. This algorithm has strong robustness and real time.
For the reduction of human face recognition rate under non-uniform illumination,Laplace filter and discrete cosine transform (DCT) gradient direction histogram (HOG) face recognition algorithm is proposed. First by Laplace filtering of face image processing,its texture feature is highlighted; Secondly in discrete cosine transform (DCT),high frequency components are effectively filtered out; Then using discrete cosine inverse transformation (IDCT) reconstruction of face image reduces its dimension; Finally through the gradient direction histogram (HOG) operator the inherent characteristics of facial image recognition are extracted and the nearest neighbor method is used. The experimental results show that the proposed algorithm under different characteristic dimension of human face recognition rate as high as 95% in the database and Yale B group of uighurs face database construction in its recognition rate is as high as 98.5%,is better than other traditional algorithm. This algorithm has strong robustness and real time.
引文
[1]Ahmad B F,Arif W M.Gabor Wavelet Based Face Recognition under Varying Lighting,Pose and Expression Conditions[C]//20152nd International Conference on Computing for Sustainable Global Development,2015:1314-1318.
[2]Mohamed A A,Yampolskiy R V.An Improved LBP Algorithm for Avatar Face Recognition[C]//XXIII International Symposium on Information,Communication and Automation Automation Technologies,2011:1-5.
[3]李伟生,王立逗,周丽芳.一种基于LTP自适应阈值的人脸识别方法[J].小型微型计算机系统,2014,9(9):2099-2103.
[4]郭金鑫,陈玮.基于HOG多特征融合与随机森林的人脸识别[J].计算机科学,2013,40(10):279-283.
[5]Dalal N,Triggs B.Histograms of Oriented Gradients for Human Detection[C]//Proceedings of the IEEE International Conference on Computer Vision and Pattern Recognition,San Diego,USA,2005:886-893.
[6]Deniz,O,Bueno,G,Salito,J,et al.Face Recognition Using Histograms of Oriented Gradients[J].Pattern Recognition Letters,2011,32(12):1598-1603.
[7]王燕,王芸芸.一种DCT和ELBP融合的人脸特征提取方法[J].计算机工程与应,2017(4):170-175.
[8]Jadhav D V,Holambe R S.Radon and Discrete Cosine Transforms Based Feature Extraction and Dimensionality Reduction Approach for Face Recognition[J].Signal Proc,2008,88(10):2604-2609.
[9]Tan Hengliang,Yang Bing,Ma Zhengming.Face Recognition Based on the Fusion of Global and Local HOG Features of Face Images[J].IET Computer Vision,2013:224-234.
[10]孙锐,陈军.基于显著性检测与HOG-NMF特征的快速行人检测方法[J].电子与信息学报,2013,35(8):1921-1925.
[11]Park W J,Kim D H.Fast Human Detection Using Selective BlockBased HOG-LBP[C]//IEEE International Conference on Image Processing,2012:601-604.
[2]Mohamed A A,Yampolskiy R V.An Improved LBP Algorithm for Avatar Face Recognition[C]//XXIII International Symposium on Information,Communication and Automation Automation Technologies,2011:1-5.
[3]李伟生,王立逗,周丽芳.一种基于LTP自适应阈值的人脸识别方法[J].小型微型计算机系统,2014,9(9):2099-2103.
[4]郭金鑫,陈玮.基于HOG多特征融合与随机森林的人脸识别[J].计算机科学,2013,40(10):279-283.
[5]Dalal N,Triggs B.Histograms of Oriented Gradients for Human Detection[C]//Proceedings of the IEEE International Conference on Computer Vision and Pattern Recognition,San Diego,USA,2005:886-893.
[6]Deniz,O,Bueno,G,Salito,J,et al.Face Recognition Using Histograms of Oriented Gradients[J].Pattern Recognition Letters,2011,32(12):1598-1603.
[7]王燕,王芸芸.一种DCT和ELBP融合的人脸特征提取方法[J].计算机工程与应,2017(4):170-175.
[8]Jadhav D V,Holambe R S.Radon and Discrete Cosine Transforms Based Feature Extraction and Dimensionality Reduction Approach for Face Recognition[J].Signal Proc,2008,88(10):2604-2609.
[9]Tan Hengliang,Yang Bing,Ma Zhengming.Face Recognition Based on the Fusion of Global and Local HOG Features of Face Images[J].IET Computer Vision,2013:224-234.
[10]孙锐,陈军.基于显著性检测与HOG-NMF特征的快速行人检测方法[J].电子与信息学报,2013,35(8):1921-1925.
[11]Park W J,Kim D H.Fast Human Detection Using Selective BlockBased HOG-LBP[C]//IEEE International Conference on Image Processing,2012:601-604.