Pub Date : 2009-12-04DOI: 10.1109/CCPR.2009.5343993
Weiling Cai, Songcan Chen, Lei Lei
In this paper, a fuzzy clustering algorithm using dependable neighbor pixels is proposed for image segmentation. In order to enhance the segmentation performance, the proposed algortihm utilizes the local statistical information to discriminate dependable neighbor pixels from undependable neighbor pixels, and then allows the labeling of the pixel to be influenced by the dependable neighbor pixels. This algorithm has two advantages: (1) the spatial information with high reliability is incorporated into the objective function so that the segmentation accuracy is guaranteed; (2) the intensity of the spatial constraints is automatically determined by the similarity meature so that the segmentation result is adaptive to the original image. The efficiency of the proposed algorithm is demonstrated by extensive segmentation experiments using both synthetic and real images.
{"title":"A Fuzzy Clustering Algorithm for Image Segmentation Using Dependable Neighbor Pixels","authors":"Weiling Cai, Songcan Chen, Lei Lei","doi":"10.1109/CCPR.2009.5343993","DOIUrl":"https://doi.org/10.1109/CCPR.2009.5343993","url":null,"abstract":"In this paper, a fuzzy clustering algorithm using dependable neighbor pixels is proposed for image segmentation. In order to enhance the segmentation performance, the proposed algortihm utilizes the local statistical information to discriminate dependable neighbor pixels from undependable neighbor pixels, and then allows the labeling of the pixel to be influenced by the dependable neighbor pixels. This algorithm has two advantages: (1) the spatial information with high reliability is incorporated into the objective function so that the segmentation accuracy is guaranteed; (2) the intensity of the spatial constraints is automatically determined by the similarity meature so that the segmentation result is adaptive to the original image. The efficiency of the proposed algorithm is demonstrated by extensive segmentation experiments using both synthetic and real images.","PeriodicalId":354468,"journal":{"name":"2009 Chinese Conference on Pattern Recognition","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121884307","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2009-12-04DOI: 10.1109/CCPR.2009.5344110
Junhua Li, Ming Li, Xiaoqin Yang
To extend GA's application, that is important to study on Genetic Algorithm under noise environment. This paper firstly described the noise environment of the GA, analyzed the effect on GA of noise; then two indexes were proposed to evaluate the performance of GA in the noisy environment, CBMPGA was proposed for the noisy optimization, the numerical experiment shows that the performance of CBMPGA is better than EGA and DCGA.
{"title":"Cluster Based Multi-Populations Genetic Algorithm in Noisy Environment","authors":"Junhua Li, Ming Li, Xiaoqin Yang","doi":"10.1109/CCPR.2009.5344110","DOIUrl":"https://doi.org/10.1109/CCPR.2009.5344110","url":null,"abstract":"To extend GA's application, that is important to study on Genetic Algorithm under noise environment. This paper firstly described the noise environment of the GA, analyzed the effect on GA of noise; then two indexes were proposed to evaluate the performance of GA in the noisy environment, CBMPGA was proposed for the noisy optimization, the numerical experiment shows that the performance of CBMPGA is better than EGA and DCGA.","PeriodicalId":354468,"journal":{"name":"2009 Chinese Conference on Pattern Recognition","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128319839","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2009-12-04DOI: 10.1109/CCPR.2009.5344038
Qiang Chen, Zexuan Ji, Quansen Sun, D. Xia
This paper presents a homogeneous patch based fuzzy c-means (FCM) clustering algorithm for brain magnetic resonance (MR) image segmentation. Currently, FCM is mainly improved by incorporating local spatial information for noise immunity. The proposed algorithm is based on image patch space, which can avoid introducing an extra control parameter for local spatial restriction. In order to decrease the edge blurring caused by local spatial restriction, the local polynomial approximation-intersection of confidence intervals (LPA-ICI) technique is used to construct the homogeneous patch. Brain MR image segmentation results indicate that the proposed algorithm is better than the other improved FCM algorithms that incorporate local spatial information, while the detail preservation need to be improved.
{"title":"Homogeneous Patch Based FCM Algorithm for Brain MR Image Segmentation","authors":"Qiang Chen, Zexuan Ji, Quansen Sun, D. Xia","doi":"10.1109/CCPR.2009.5344038","DOIUrl":"https://doi.org/10.1109/CCPR.2009.5344038","url":null,"abstract":"This paper presents a homogeneous patch based fuzzy c-means (FCM) clustering algorithm for brain magnetic resonance (MR) image segmentation. Currently, FCM is mainly improved by incorporating local spatial information for noise immunity. The proposed algorithm is based on image patch space, which can avoid introducing an extra control parameter for local spatial restriction. In order to decrease the edge blurring caused by local spatial restriction, the local polynomial approximation-intersection of confidence intervals (LPA-ICI) technique is used to construct the homogeneous patch. Brain MR image segmentation results indicate that the proposed algorithm is better than the other improved FCM algorithms that incorporate local spatial information, while the detail preservation need to be improved.","PeriodicalId":354468,"journal":{"name":"2009 Chinese Conference on Pattern Recognition","volume":"32 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129046671","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2009-12-04DOI: 10.1109/CCPR.2009.5344013
G. Zhong, Xinwen Hou, Cheng-Lin Liu
In many areas of pattern recognition and machine learning, low dimensional data are often embedded in a high dimensional space. There have been many dimensionality reduction and manifold learning methods to discover the low dimensional representation from high dimensional data. Locality based manifold learning methods often rely on a distance metric between neighboring points. In this paper, we propose a new distance metric named relative distance, which is learned from the data and can better reflect the relative density. Combining the relative distance with Laplacian Eigenmaps (LE), we obtain a new algorithm called Relative Distance-based Laplacian Eigenmaps (RDLE) for nonlinear dimensionality reduction. Based on two different definitions of the relative distance, we give two variations of the RDLE. For efficient projection of out-of-sample data, we also present the linear version of RDLE, LRDLE. Experimental results on toy problems and real-world data demonstrate the effectiveness of our methods.
{"title":"Relative Distance-Based Laplacian Eigenmaps","authors":"G. Zhong, Xinwen Hou, Cheng-Lin Liu","doi":"10.1109/CCPR.2009.5344013","DOIUrl":"https://doi.org/10.1109/CCPR.2009.5344013","url":null,"abstract":"In many areas of pattern recognition and machine learning, low dimensional data are often embedded in a high dimensional space. There have been many dimensionality reduction and manifold learning methods to discover the low dimensional representation from high dimensional data. Locality based manifold learning methods often rely on a distance metric between neighboring points. In this paper, we propose a new distance metric named relative distance, which is learned from the data and can better reflect the relative density. Combining the relative distance with Laplacian Eigenmaps (LE), we obtain a new algorithm called Relative Distance-based Laplacian Eigenmaps (RDLE) for nonlinear dimensionality reduction. Based on two different definitions of the relative distance, we give two variations of the RDLE. For efficient projection of out-of-sample data, we also present the linear version of RDLE, LRDLE. Experimental results on toy problems and real-world data demonstrate the effectiveness of our methods.","PeriodicalId":354468,"journal":{"name":"2009 Chinese Conference on Pattern Recognition","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129197318","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2009-12-04DOI: 10.1109/CCPR.2009.5344065
Yan Zhang, Fanglin Wang, Shengyang Yu
For multitarget tracking problems, occlusions between targets are quite tough tasks. We present a novel algorithm to solve such problems. For the two targets in occlusions, Fukunaga-Koontz transform is exploited to achieve the projection matrix, with which the two targets are projected into a low dimensional space where they are quite distinguishing. To solve the problem of the change of target appearance, the eigenspace model is used as the probabilistic observation model, with which the algorithm can learn the changes of the target appearance online. These two procedures are evaluated in the particle filter based tracking framework. Experimental results demonstrated the effectiveness of our algorithm.
{"title":"Use Fukunaga-Koontz Transform to Solve Occlusion Problems in Multitarget Tracking","authors":"Yan Zhang, Fanglin Wang, Shengyang Yu","doi":"10.1109/CCPR.2009.5344065","DOIUrl":"https://doi.org/10.1109/CCPR.2009.5344065","url":null,"abstract":"For multitarget tracking problems, occlusions between targets are quite tough tasks. We present a novel algorithm to solve such problems. For the two targets in occlusions, Fukunaga-Koontz transform is exploited to achieve the projection matrix, with which the two targets are projected into a low dimensional space where they are quite distinguishing. To solve the problem of the change of target appearance, the eigenspace model is used as the probabilistic observation model, with which the algorithm can learn the changes of the target appearance online. These two procedures are evaluated in the particle filter based tracking framework. Experimental results demonstrated the effectiveness of our algorithm.","PeriodicalId":354468,"journal":{"name":"2009 Chinese Conference on Pattern Recognition","volume":"48 3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114228844","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2009-12-04DOI: 10.1109/CCPR.2009.5344095
S. Liu, J. Wang, Hong Wang, Ling Zou
Aiming at the limitation of traditional graph-theory clustering method in the process of image segmentation, a new segmentation approach is proposed, which uses fuzzy similarity relationship to weight the edges while a complete graph is constituted. And fuzzy maximum spanning tree is used to clustering. Thus the traditional graph-theory clustering method is improved as the fuzzy graph-theory clustering method. Use the local mean and local variance to construct bivector, define the pixel's local mean and variance vector., then get the fuxxy similarity relationship of each pixel in the picture sequence. Experiments are conducted on two real pictures by MATLAB. Results show that different effects can be get by changing the parameter. And the flexibility is better than other contrast methods'.
{"title":"A New Segmentation Approach Based on Fuzzy Graph-Theory Clustering","authors":"S. Liu, J. Wang, Hong Wang, Ling Zou","doi":"10.1109/CCPR.2009.5344095","DOIUrl":"https://doi.org/10.1109/CCPR.2009.5344095","url":null,"abstract":"Aiming at the limitation of traditional graph-theory clustering method in the process of image segmentation, a new segmentation approach is proposed, which uses fuzzy similarity relationship to weight the edges while a complete graph is constituted. And fuzzy maximum spanning tree is used to clustering. Thus the traditional graph-theory clustering method is improved as the fuzzy graph-theory clustering method. Use the local mean and local variance to construct bivector, define the pixel's local mean and variance vector., then get the fuxxy similarity relationship of each pixel in the picture sequence. Experiments are conducted on two real pictures by MATLAB. Results show that different effects can be get by changing the parameter. And the flexibility is better than other contrast methods'.","PeriodicalId":354468,"journal":{"name":"2009 Chinese Conference on Pattern Recognition","volume":"16 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121619874","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2009-12-04DOI: 10.1109/CCPR.2009.5344126
Yuan-Cheng Xie, Jing-yu Yang
AdaBoost has been the representation of ensemble learning algorithm because of its excellent performance. However, due to its longtime training, AdaBoost was complained about by people and this defect limits the practical application. Bagging is a rapid method of training and supports for parallel computing. One of important factors that can affect the performance of ensemble learning is the diversity of component learners. Based on this view, a new algorithm using clustering and Boosting to prune Bagging ensembles is proposed in this paper. Its learning efficiency is close to Bagging and its performance is close to AdaBoost. Furthermore, this new algorithm can detect noisy data from original samples based on cascade technique, and a better result of noise detection can be acquired.
{"title":"Using Boosting and Clustering to Prune Bagging and Detect Noisy Data","authors":"Yuan-Cheng Xie, Jing-yu Yang","doi":"10.1109/CCPR.2009.5344126","DOIUrl":"https://doi.org/10.1109/CCPR.2009.5344126","url":null,"abstract":"AdaBoost has been the representation of ensemble learning algorithm because of its excellent performance. However, due to its longtime training, AdaBoost was complained about by people and this defect limits the practical application. Bagging is a rapid method of training and supports for parallel computing. One of important factors that can affect the performance of ensemble learning is the diversity of component learners. Based on this view, a new algorithm using clustering and Boosting to prune Bagging ensembles is proposed in this paper. Its learning efficiency is close to Bagging and its performance is close to AdaBoost. Furthermore, this new algorithm can detect noisy data from original samples based on cascade technique, and a better result of noise detection can be acquired.","PeriodicalId":354468,"journal":{"name":"2009 Chinese Conference on Pattern Recognition","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127776084","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2009-12-04DOI: 10.1109/CCPR.2009.5344108
Pei-xuan Chen, G. Feng
Most of the existing Content Based Image Retrieval algorithms are implemented in spatial domain. In order to save the time in images decompression, a novel image retrieval based on DCT dominant color and texture features in compressed domain is proposed. The mean value and variance of the DCT coefficients in DCT sub-blocks are used to describe image features. The mean values of those DCT sub-blocks with smaller variances than the threshold are regarded as the dominant colors and used to construct the image index, and we extract the texture feature from the DCT sub-blocks with larger variance than the threshold. The experimental results demonstrate the algorithm has good performance.
{"title":"Image Retrieval Based on Dominant Color and Texture Features in DCT Domain","authors":"Pei-xuan Chen, G. Feng","doi":"10.1109/CCPR.2009.5344108","DOIUrl":"https://doi.org/10.1109/CCPR.2009.5344108","url":null,"abstract":"Most of the existing Content Based Image Retrieval algorithms are implemented in spatial domain. In order to save the time in images decompression, a novel image retrieval based on DCT dominant color and texture features in compressed domain is proposed. The mean value and variance of the DCT coefficients in DCT sub-blocks are used to describe image features. The mean values of those DCT sub-blocks with smaller variances than the threshold are regarded as the dominant colors and used to construct the image index, and we extract the texture feature from the DCT sub-blocks with larger variance than the threshold. The experimental results demonstrate the algorithm has good performance.","PeriodicalId":354468,"journal":{"name":"2009 Chinese Conference on Pattern Recognition","volume":"35 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127927954","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2009-12-04DOI: 10.1109/CCPR.2009.5344147
Dong-sheng Zhang, Chao Ji
Aim at fuzzy clustering method based on transitive closure, the result of λ-level-set is arbitrary. To avoid the wrong clustering result causing by improper λ value, one scheme is presented, which makes use of dynamic clustering method to get all meaningful possibility of clustering, and then compares inner-class distance and inter-class distance respectively, and introduces F test method using in mathematical statistics to decide the best clustering scheme. Simulation result shows that the way is concise, effective and credible.
{"title":"Study on Dynamic Fuzzy Clustering and the Best Effect of Clustering","authors":"Dong-sheng Zhang, Chao Ji","doi":"10.1109/CCPR.2009.5344147","DOIUrl":"https://doi.org/10.1109/CCPR.2009.5344147","url":null,"abstract":"Aim at fuzzy clustering method based on transitive closure, the result of λ-level-set is arbitrary. To avoid the wrong clustering result causing by improper λ value, one scheme is presented, which makes use of dynamic clustering method to get all meaningful possibility of clustering, and then compares inner-class distance and inter-class distance respectively, and introduces F test method using in mathematical statistics to decide the best clustering scheme. Simulation result shows that the way is concise, effective and credible.","PeriodicalId":354468,"journal":{"name":"2009 Chinese Conference on Pattern Recognition","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121765007","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2009-12-04DOI: 10.1109/CCPR.2009.5344136
Zhihui Lai, Zhong Jin, Jian Yang
In this paper, we propose a novel supervised learning method called Global Sparse Representation Projections (GSRP) for linear dimensionality reduction. GSRP can be viewed as a combiner of sparse representation and manifold learning. But differing from the recent manifold learning methods such as Local Preserving Projections (LPP), GSRP introduces the global sparse representation information into the objective function. Since sparse representation can implicitly employ the "local" structure of the data by imposing the sparsity prior, we take advantages of this property to characterize the local structure. By combining the local interclass neighborhood relationship and sparse representation information, GSRP aims to preserve the sparse reconstructive relationship of the data and simultaneously maximize the interclass separability. Comprehensive comparison and extensive experiments show that GSRP achieves higher recognition rates than the state-of-the-art techniques such as LPP and Sparsity Preserving Projections (SPP).
{"title":"Global Sparse Representation Projections for Feature Extraction and Classification","authors":"Zhihui Lai, Zhong Jin, Jian Yang","doi":"10.1109/CCPR.2009.5344136","DOIUrl":"https://doi.org/10.1109/CCPR.2009.5344136","url":null,"abstract":"In this paper, we propose a novel supervised learning method called Global Sparse Representation Projections (GSRP) for linear dimensionality reduction. GSRP can be viewed as a combiner of sparse representation and manifold learning. But differing from the recent manifold learning methods such as Local Preserving Projections (LPP), GSRP introduces the global sparse representation information into the objective function. Since sparse representation can implicitly employ the \"local\" structure of the data by imposing the sparsity prior, we take advantages of this property to characterize the local structure. By combining the local interclass neighborhood relationship and sparse representation information, GSRP aims to preserve the sparse reconstructive relationship of the data and simultaneously maximize the interclass separability. Comprehensive comparison and extensive experiments show that GSRP achieves higher recognition rates than the state-of-the-art techniques such as LPP and Sparsity Preserving Projections (SPP).","PeriodicalId":354468,"journal":{"name":"2009 Chinese Conference on Pattern Recognition","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131142579","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：