一种基于局部排序的约束稀疏编码的图像分类方法
|
摘要
图像分类作为计算机视觉分析领域一个重要的研究方向,其分类性能很大程度上取决于图像的特征表示.为了能够更好地进行图像分类,本文提出了一种基于局部约束稀疏编码的神经气算法(Neural Gas based Locality-constrained Sparse Coding,NGLSC)用来实现图像分类.引入局部排序适配器作为距离正则化约束项已经应用在神经气(Neural Gas,NG)的算法矢量量化中,旨在通过软竞争学习算法来弥补K均值聚类(K-means)算法的不足.在稀疏编码阶段此算法可求解得到封闭解.此外,字典更新一般由目标函数的误差项来决定,已有一些经典的算法采用这种方式更新字典.本文使用ORL数据库和COIL20数据库将所提出算法和现有算法局部约束线性编码(Locality-constrained Linear Coding,LLC),脸元数据学习方法(Metaface Learning,MFL)进行比较.实验结果证明本文所提出的算法在图像分类上准确率可达95%以上.可以看出,本文为计算机视觉图像分类工作提供了一种有价值的解决思路.
Image classification is an important research direction in the field of computer vision analysis.The key to classification depends largely on the feature representation of the image.In order to be able to classify images better,this paper presents a locality-constrained sparse coding for image classification by introducing ranking locality adaptor as distance regularization.The proposed ranking locality adaptor has previously been used in Neural Gas(NG) method for vector quantization,which originally remedies the K-means by using this soft-competitive learning scheme.In the new proposed method,a closed-form solution can be derived at sparse coding step.In addition,dictionary updates are generally determined by the error term of the objective function.Some classical algorithms have used this method to update the dictionary.This paper uses the ORL database and COIL20 database to compare the proposed algorithm with the existing algorithm Locality-constrained Linear Coding,and Metaface Learning algorithm.The experimental results show that the proposed algorithm has an accuracy of more than 95% in image classification and has stronger performance than the current excellent algorithms.In addition,the recognition rate of the algorithm does not change greatly with the change of the data feature dimension.It can be seen that this paper provides a valuable solution for the classification of computer vision images.
Image classification is an important research direction in the field of computer vision analysis.The key to classification depends largely on the feature representation of the image.In order to be able to classify images better,this paper presents a locality-constrained sparse coding for image classification by introducing ranking locality adaptor as distance regularization.The proposed ranking locality adaptor has previously been used in Neural Gas(NG) method for vector quantization,which originally remedies the K-means by using this soft-competitive learning scheme.In the new proposed method,a closed-form solution can be derived at sparse coding step.In addition,dictionary updates are generally determined by the error term of the objective function.Some classical algorithms have used this method to update the dictionary.This paper uses the ORL database and COIL20 database to compare the proposed algorithm with the existing algorithm Locality-constrained Linear Coding,and Metaface Learning algorithm.The experimental results show that the proposed algorithm has an accuracy of more than 95% in image classification and has stronger performance than the current excellent algorithms.In addition,the recognition rate of the algorithm does not change greatly with the change of the data feature dimension.It can be seen that this paper provides a valuable solution for the classification of computer vision images.
引文
[1] Turk M,Pentland A.Eigenfaces for recognition[J].J Cogn Neurosci,1991,3(1):71-86.
[2] Roweis S T,Saul L K.Nonlinear dimensionality reduction by locally linear embedding[J].Science,2000,290(5500):2323-2326.
[3] Wright J,Yang A Y,Ganesh A,et al.Robust face recognition via sparse representation[J].IEEE Transactions on Pattern Analysis & Machine Intelligence,2009,31(2):210-227.
[4] Olshausen B A,Field D J.Emergence of simple-cell receptive field properties by learning a sparse code for natural images[J].Nature,1996,381(6583):607-609.
[5] Yang M,Zhang L,Yang J,et al.Metaface learning for sparse representation based face recognition[A].IEEE International Conference on Image Processing[C].IEEE,2010.1601-1604.
[6] Wang J,Yang J,Yu K,et al.Locality-constrained linear coding for image classification[A].Computer Vision and Pattern Recognition[C].IEEE,2010.3360-3367.
[7] Wei C P,Chao Y W,Yeh Y R,et al.Locality-sensitive dictionary learning for sparse representation based classification[J].Pattern Recognition,2013,46(5):1277-1287.
[8] Martinetz T M,Berkovich S G,Schulten K J.`Neural-gas' network for vector quantization and its application to time-series prediction[J].IEEE Trans Neural Netw,1993,4(4):558-569.
[9] Engan K,Aase S O,Husoy J H.Frame based signal compression using method of optimal directions (MOD)[A].IEEE International Symposium on Circuits and Systems[C].IEEE,1999.1-4.
[10] Aharon M,Elad M,Bruckstein A.K-SVD:An algorithm for designing overcomplete dictionaries for sparse representation[J].IEEE Transactions on Signal Processing,2006,54(11):4311-4322.
[11] Rubinstein R,Zibulevsky M,Elad M.Efficient implementation of the K-SVD algorithm using batch orthogonal matching pursuit[J].CS Technion,2008,40.
[12] Samaria F S,Harter A C.Parameterisation of a stochastic model for human face identification[A].Proceedings of the Second IEEE Workshop on Applications of Computer Vision[C].IEEE,1994.138-142.
[13] 邓贞宙,段志文,刘晶晶,谢庆国.基于一种新型扫描模式的不完全数据CT重建[J].核电子学与探测技术,2011,31(1):53-57.Z Deng,Z Duan,J Liu,Q Xie.Incomplete data CT image reconstruction based on a new scanning mode[J].Nuclear Electronics & Detection Technology,2011,31(1):53-57.(in Chinese)
[14] 邓贞宙,段志文,谢庆国.基于微元法的高精度系统响应矩阵建模[J].核电子学与探测技术,2011,31(10):1125-1130.Z Deng,Z Duan,Q Xie.System response matrix modeling with high accuracy via infinitesimal analysis[J].Nuclear Electronics & Detection Technology,2011,31(10):1125-1130.(in Chinese)
[2] Roweis S T,Saul L K.Nonlinear dimensionality reduction by locally linear embedding[J].Science,2000,290(5500):2323-2326.
[3] Wright J,Yang A Y,Ganesh A,et al.Robust face recognition via sparse representation[J].IEEE Transactions on Pattern Analysis & Machine Intelligence,2009,31(2):210-227.
[4] Olshausen B A,Field D J.Emergence of simple-cell receptive field properties by learning a sparse code for natural images[J].Nature,1996,381(6583):607-609.
[5] Yang M,Zhang L,Yang J,et al.Metaface learning for sparse representation based face recognition[A].IEEE International Conference on Image Processing[C].IEEE,2010.1601-1604.
[6] Wang J,Yang J,Yu K,et al.Locality-constrained linear coding for image classification[A].Computer Vision and Pattern Recognition[C].IEEE,2010.3360-3367.
[7] Wei C P,Chao Y W,Yeh Y R,et al.Locality-sensitive dictionary learning for sparse representation based classification[J].Pattern Recognition,2013,46(5):1277-1287.
[8] Martinetz T M,Berkovich S G,Schulten K J.`Neural-gas' network for vector quantization and its application to time-series prediction[J].IEEE Trans Neural Netw,1993,4(4):558-569.
[9] Engan K,Aase S O,Husoy J H.Frame based signal compression using method of optimal directions (MOD)[A].IEEE International Symposium on Circuits and Systems[C].IEEE,1999.1-4.
[10] Aharon M,Elad M,Bruckstein A.K-SVD:An algorithm for designing overcomplete dictionaries for sparse representation[J].IEEE Transactions on Signal Processing,2006,54(11):4311-4322.
[11] Rubinstein R,Zibulevsky M,Elad M.Efficient implementation of the K-SVD algorithm using batch orthogonal matching pursuit[J].CS Technion,2008,40.
[12] Samaria F S,Harter A C.Parameterisation of a stochastic model for human face identification[A].Proceedings of the Second IEEE Workshop on Applications of Computer Vision[C].IEEE,1994.138-142.
[13] 邓贞宙,段志文,刘晶晶,谢庆国.基于一种新型扫描模式的不完全数据CT重建[J].核电子学与探测技术,2011,31(1):53-57.Z Deng,Z Duan,J Liu,Q Xie.Incomplete data CT image reconstruction based on a new scanning mode[J].Nuclear Electronics & Detection Technology,2011,31(1):53-57.(in Chinese)
[14] 邓贞宙,段志文,谢庆国.基于微元法的高精度系统响应矩阵建模[J].核电子学与探测技术,2011,31(10):1125-1130.Z Deng,Z Duan,Q Xie.System response matrix modeling with high accuracy via infinitesimal analysis[J].Nuclear Electronics & Detection Technology,2011,31(10):1125-1130.(in Chinese)