一种改进模糊C均值聚类的图像标注方法
|
摘要
本文主要利用图像底层特征以及图像标签的语义信息对图像进行自动标注,在此基础上提出了改进模糊C均值(FCM)聚类的标注方法.首先结合图像特征以及同类、异类样本间的关系信息,融合聚类中心之间的距离,改善了算法中距离测度较为单一的问题.在目标函数中将传统的距离测度改为同类样本距离与异类样本距离之差,体现了同类样本的密度和异类样本的稀疏程度,提高了标注准确率.然后使用改进后的算法对每类图像进行聚类,计算待标注图像到各个聚类中心的平均距离来判断其类别.之后计算图像到各个子类的聚类中心的距离,并统计所属类内的标注词即为图像的标注词.利用Corel5K和iaprtc12来验证算法的可行性,通过实验对比不同测度以及分析不同标注模型的结果,表明该方法有效的提高了标注准确率.
This paper mainly uses the underlying information of images and the semantic features of image tags to automatically annotate images. On the basis of this,we propose an improved fuzzy C means(Fuzzy C-means) clustering annotation method. Firstly,the distance between the clustering centers is combined with the relationship between the identical samples and the similar heterogeneous samples,which improves the problem of the distance measure in the algorithm. In the objective function,the traditional distance measure is changed to the distance between the similar sample and the heterogeneous sample,which reflects the density of the similar sample and the degree of discretization of the heterogeneous sample,and improves the accuracy of annotating. Then,the improved algorithm is used to cluster FCMof each image,and then the average distance of the image to each clustering centers is calculated to determine the category of the image. Then,the distance between the image and the clustering centers of each subclass is calculated,and the tagged words in the genus are calculated as the annotated words of the image. Corel5 K and iaprtc12 are used to verify the feasibility of the test,The results of different measurement and different annotation models were compared by experiment,The experiment shows that the method can effectively improve the rate of Labeling accuracy.
This paper mainly uses the underlying information of images and the semantic features of image tags to automatically annotate images. On the basis of this,we propose an improved fuzzy C means(Fuzzy C-means) clustering annotation method. Firstly,the distance between the clustering centers is combined with the relationship between the identical samples and the similar heterogeneous samples,which improves the problem of the distance measure in the algorithm. In the objective function,the traditional distance measure is changed to the distance between the similar sample and the heterogeneous sample,which reflects the density of the similar sample and the degree of discretization of the heterogeneous sample,and improves the accuracy of annotating. Then,the improved algorithm is used to cluster FCMof each image,and then the average distance of the image to each clustering centers is calculated to determine the category of the image. Then,the distance between the image and the clustering centers of each subclass is calculated,and the tagged words in the genus are calculated as the annotated words of the image. Corel5 K and iaprtc12 are used to verify the feasibility of the test,The results of different measurement and different annotation models were compared by experiment,The experiment shows that the method can effectively improve the rate of Labeling accuracy.
引文
[1]Luo J,Savakis A.Indoor vs outdoor classification of consumer photographs using low-level and semantic features[C].International Conference on Image Processing,Proceedings,IEEE,2001:745-748.
[2]Cusano C,Ciocca G,Schettini R.Image annotation using SVM[C].Electronic Imaging,International Society for Optics and Photonics,2003:330-338.
[3]Wang J Z,Li J.Learning-based linguistic indexing of pictures with2-d M HM M s[C].ACM M ultimedia,2002:436-445.
[4]Ghoshal A,Ircing P,Khudanpur S.Hidden Markov models for automatic annotation and content-based retrieval of images and video[C].International ACM SIGIR Conference on Research and Development in Information Retrieval,ACM,2005:544-551.
[5]Boutell M R,Luo J,Shen X,et al.Learning multi-label scene classification[J].Pattern Recognition,2004,37(9):1757-1771.
[6]Tsai C F,Mcgarry K,Tait J.CLAIRE:a modular support vector image indexing and classification system[J].Acm Transactions on Information Systems,2006,24(3):353-379.
[7]Makadia A,Pavlovic V,Kumar S.Baselines for image annotation[J].International Journal of Computer Vision,2010,90(1):88-105.
[8]Duygulu P,Barnard K,Freitas J F G D,et al.Object recognition as machine translation:learning a lexicon for a fixed image vocabulary[C].European Conference on Computer Vision,Springer-Verlag,2002:97-112.
[9]Feng S L,Manmatha R,Lavrenko V.Multiple bernoulli relevance models for image and video annotation[C].Computer Vision and Pattern Recognition,CVPR,Proceedings of the 2004 IEEE Computer Society Conference on,IEEE,2004:1002-1009.
[10]Monay F,Gatica-Perez D.On image auto-annotation with latent space models[C].Proc.acm Int.conf.on Multimedia,2003:275-278.
[11]Monay F,Gaticaperez D.PLSA-based image auto-annotation:constraining the latent space[C].Proc.acm Int.conf.on M ultimedia,2004:348-351.
[12]Li Zhi-xin,Shi Zhi-ping,Li Zhi-qing.Automatic annotation of images based on semantic topics[J].Journal of Softw are,2011,22(4):801-812.
[13]Lu Han-qing,Liu Jing.Automatic image annotation based on graph learning[J].Chinese Journal of Computers,2008,31(9):1629-1639.
[14]Yuan Ying,Shao Jian,Wu Fei,et al.Image annotation based on sparse effect and multi kernel learning[J].Journal of Softw are,2012,23(9):2500-2509.
[15]Bao Hong,Xu Guang-mei,Feng Song-he,et al.Research progress of automatic image annotation technology[J].Computer Science,2011,38(7):35-40.
[16]Liu Kai,Zhang Li-min,Sun Yong-wei,et al.Automatic image annotation algorithm using depth Boltzmann machine and canonical correlation analysis[J].Journal of Xi'an Jiaotong University,2015,49(6):33-38.
[17]Zhang Xiao-chun.Image annotation based on maximum probability method and nearest neighbor criterion[D].Nanjing:Nanjing University of Science and Technology,2014.
[18]Liu Lu,Wu Cheng-mao.Fuzzy C-mean clustering segmentation algorithm based on intra class distance[J].Computer Engineering and Design,2016,37(6):1626-1631.
[12]李志欣,施智平,李志清,等.融合语义主题的图像自动标注[J].软件学报,2011,22(4):801-812.
[13]卢汉清,刘静.基于图学习的自动图像标注[J].计算机学报,2008,31(9):1629-1639.
[14]袁莹,邵健,吴飞,等.结合组稀疏效应和多核学习的图像标注[J].软件学报,2012,23(9):2500-2509.
[15]鲍泓,徐光美,冯松鹤,等.自动图像标注技术研究进展[J].计算机科学,2011,38(7):35-40.
[16]刘凯,张立民,孙永威,等.利用深度玻尔兹曼机与典型相关分析的自动图像标注算法[J].西安交通大学学报,2015,49(6):33-38.
[17]张晓春.最大概率方法与最近邻准则下的图像标注[D].南京:南京理工大学,2014.
[18]刘璐,吴成茂.基于类内类间距离的模糊C-均值聚类分割算法[J].计算机工程与设计,2016,37(6):1626-1631.
[2]Cusano C,Ciocca G,Schettini R.Image annotation using SVM[C].Electronic Imaging,International Society for Optics and Photonics,2003:330-338.
[3]Wang J Z,Li J.Learning-based linguistic indexing of pictures with2-d M HM M s[C].ACM M ultimedia,2002:436-445.
[4]Ghoshal A,Ircing P,Khudanpur S.Hidden Markov models for automatic annotation and content-based retrieval of images and video[C].International ACM SIGIR Conference on Research and Development in Information Retrieval,ACM,2005:544-551.
[5]Boutell M R,Luo J,Shen X,et al.Learning multi-label scene classification[J].Pattern Recognition,2004,37(9):1757-1771.
[6]Tsai C F,Mcgarry K,Tait J.CLAIRE:a modular support vector image indexing and classification system[J].Acm Transactions on Information Systems,2006,24(3):353-379.
[7]Makadia A,Pavlovic V,Kumar S.Baselines for image annotation[J].International Journal of Computer Vision,2010,90(1):88-105.
[8]Duygulu P,Barnard K,Freitas J F G D,et al.Object recognition as machine translation:learning a lexicon for a fixed image vocabulary[C].European Conference on Computer Vision,Springer-Verlag,2002:97-112.
[9]Feng S L,Manmatha R,Lavrenko V.Multiple bernoulli relevance models for image and video annotation[C].Computer Vision and Pattern Recognition,CVPR,Proceedings of the 2004 IEEE Computer Society Conference on,IEEE,2004:1002-1009.
[10]Monay F,Gatica-Perez D.On image auto-annotation with latent space models[C].Proc.acm Int.conf.on Multimedia,2003:275-278.
[11]Monay F,Gaticaperez D.PLSA-based image auto-annotation:constraining the latent space[C].Proc.acm Int.conf.on M ultimedia,2004:348-351.
[12]Li Zhi-xin,Shi Zhi-ping,Li Zhi-qing.Automatic annotation of images based on semantic topics[J].Journal of Softw are,2011,22(4):801-812.
[13]Lu Han-qing,Liu Jing.Automatic image annotation based on graph learning[J].Chinese Journal of Computers,2008,31(9):1629-1639.
[14]Yuan Ying,Shao Jian,Wu Fei,et al.Image annotation based on sparse effect and multi kernel learning[J].Journal of Softw are,2012,23(9):2500-2509.
[15]Bao Hong,Xu Guang-mei,Feng Song-he,et al.Research progress of automatic image annotation technology[J].Computer Science,2011,38(7):35-40.
[16]Liu Kai,Zhang Li-min,Sun Yong-wei,et al.Automatic image annotation algorithm using depth Boltzmann machine and canonical correlation analysis[J].Journal of Xi'an Jiaotong University,2015,49(6):33-38.
[17]Zhang Xiao-chun.Image annotation based on maximum probability method and nearest neighbor criterion[D].Nanjing:Nanjing University of Science and Technology,2014.
[18]Liu Lu,Wu Cheng-mao.Fuzzy C-mean clustering segmentation algorithm based on intra class distance[J].Computer Engineering and Design,2016,37(6):1626-1631.
[12]李志欣,施智平,李志清,等.融合语义主题的图像自动标注[J].软件学报,2011,22(4):801-812.
[13]卢汉清,刘静.基于图学习的自动图像标注[J].计算机学报,2008,31(9):1629-1639.
[14]袁莹,邵健,吴飞,等.结合组稀疏效应和多核学习的图像标注[J].软件学报,2012,23(9):2500-2509.
[15]鲍泓,徐光美,冯松鹤,等.自动图像标注技术研究进展[J].计算机科学,2011,38(7):35-40.
[16]刘凯,张立民,孙永威,等.利用深度玻尔兹曼机与典型相关分析的自动图像标注算法[J].西安交通大学学报,2015,49(6):33-38.
[17]张晓春.最大概率方法与最近邻准则下的图像标注[D].南京:南京理工大学,2014.
[18]刘璐,吴成茂.基于类内类间距离的模糊C-均值聚类分割算法[J].计算机工程与设计,2016,37(6):1626-1631.