Pub Date : 2011-11-01DOI: 10.1109/ACPR.2011.6166601
Shizhun Yang, Chenping Hou, Yi Wu
Traditional data mining and machine learning technologies may fail when the training data and the testing data are drawn from different feature spaces and different distributions. Transfer learning, which uses the data from source domain and target domain, can tackle this problem. In this paper, we propose an improved Graph-based Model for Transfer learning (GM-Transfer). We construct a tripartite graph to represent the transfer learning problem and model the relations between the source domain data and the target domain data more efficiently. By learning the informational graph, the knowledge from the source domain data can be transferred to help improve the learning efficiency on the target domain data. Experiments show the effectiveness of our algorithm.
{"title":"GM-transfer: Graph-based model for transfer learning","authors":"Shizhun Yang, Chenping Hou, Yi Wu","doi":"10.1109/ACPR.2011.6166601","DOIUrl":"https://doi.org/10.1109/ACPR.2011.6166601","url":null,"abstract":"Traditional data mining and machine learning technologies may fail when the training data and the testing data are drawn from different feature spaces and different distributions. Transfer learning, which uses the data from source domain and target domain, can tackle this problem. In this paper, we propose an improved Graph-based Model for Transfer learning (GM-Transfer). We construct a tripartite graph to represent the transfer learning problem and model the relations between the source domain data and the target domain data more efficiently. By learning the informational graph, the knowledge from the source domain data can be transferred to help improve the learning efficiency on the target domain data. Experiments show the effectiveness of our algorithm.","PeriodicalId":287232,"journal":{"name":"The First Asian Conference on Pattern Recognition","volume":"251 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116209243","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 : 2011-11-01DOI: 10.1109/ACPR.2011.6166615
Chuan-Xian Ren, D. Dai, Hong Yan
We present a novel classification method formulating an objective model by ℓ2;1-norm based regression. The ℓ2;1-norm based loss function is robust to outliers or the large variations within given data, and the ℓ2;1-norm regularization term selects correlated samples across the whole training set with grouped sparsity. This constrained optimization problem can be efficiently solved by an iterative procedure. Several benchmark data sets including facial images and gene expression data are used for evaluating the robustness and effectiveness of the new proposed algorithm, and the results show the competitive performance.
{"title":"ℓ2;1-norm based Regression for Classification","authors":"Chuan-Xian Ren, D. Dai, Hong Yan","doi":"10.1109/ACPR.2011.6166615","DOIUrl":"https://doi.org/10.1109/ACPR.2011.6166615","url":null,"abstract":"We present a novel classification method formulating an objective model by ℓ2;1-norm based regression. The ℓ2;1-norm based loss function is robust to outliers or the large variations within given data, and the ℓ2;1-norm regularization term selects correlated samples across the whole training set with grouped sparsity. This constrained optimization problem can be efficiently solved by an iterative procedure. Several benchmark data sets including facial images and gene expression data are used for evaluating the robustness and effectiveness of the new proposed algorithm, and the results show the competitive performance.","PeriodicalId":287232,"journal":{"name":"The First Asian Conference on Pattern Recognition","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121563807","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 : 2011-11-01DOI: 10.1109/ACPR.2011.6166566
Yi C. Feng, Lei Huang, Chang-ping Liu
Palmprint recognition has attracted much attention in recent years. Many algorithms based texture coding achieve high accuracy. However they are still sensitive to local unsteady region introduced by variations of hand pose and other conditions. In this paper we proposed a novel feature extraction algorithm, namely binary contrast context vector (BCCV), to represent multiple contrast distribution for a local region. Due to forming the local contrast value into a binary vector, contrast context could be used to match more effectively. Furthermore, by using BCCV we apply an adaptive threshold to mask the stable local region before matching. Our experiment results on public palmprint database shows that the proposed BCCV achieves lower equal error rate (EER) than other two state-of-the-art approaches.
{"title":"Palmprint verification using binary contrast context vector","authors":"Yi C. Feng, Lei Huang, Chang-ping Liu","doi":"10.1109/ACPR.2011.6166566","DOIUrl":"https://doi.org/10.1109/ACPR.2011.6166566","url":null,"abstract":"Palmprint recognition has attracted much attention in recent years. Many algorithms based texture coding achieve high accuracy. However they are still sensitive to local unsteady region introduced by variations of hand pose and other conditions. In this paper we proposed a novel feature extraction algorithm, namely binary contrast context vector (BCCV), to represent multiple contrast distribution for a local region. Due to forming the local contrast value into a binary vector, contrast context could be used to match more effectively. Furthermore, by using BCCV we apply an adaptive threshold to mask the stable local region before matching. Our experiment results on public palmprint database shows that the proposed BCCV achieves lower equal error rate (EER) than other two state-of-the-art approaches.","PeriodicalId":287232,"journal":{"name":"The First Asian Conference on Pattern Recognition","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132306158","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 : 2011-11-01DOI: 10.1109/ACPR.2011.6166542
Pin-Ching Su, Hwann-Tzong Chen, Koichi Ito, T. Aoki
We present a new approach to the problem of grouping similar scene images. The proposed method characterizes both the global feature layout and the local oriented edge responses of an image, and provides a translation-invariant similarity measure to compare scene images. Our method is effective in generating initial clustering results for applications that require extensive local-feature matching on unorganized image collections, such as large-scale 3D reconstruction and scene completion. The advantage of our method is that it can estimate image similarity via integrating global and local information. The experimental evaluations on various image datasets show that our method is able to approximate well the similarities derived from local-feature matching with a lower computational cost.
{"title":"Translation-invariant scene grouping","authors":"Pin-Ching Su, Hwann-Tzong Chen, Koichi Ito, T. Aoki","doi":"10.1109/ACPR.2011.6166542","DOIUrl":"https://doi.org/10.1109/ACPR.2011.6166542","url":null,"abstract":"We present a new approach to the problem of grouping similar scene images. The proposed method characterizes both the global feature layout and the local oriented edge responses of an image, and provides a translation-invariant similarity measure to compare scene images. Our method is effective in generating initial clustering results for applications that require extensive local-feature matching on unorganized image collections, such as large-scale 3D reconstruction and scene completion. The advantage of our method is that it can estimate image similarity via integrating global and local information. The experimental evaluations on various image datasets show that our method is able to approximate well the similarities derived from local-feature matching with a lower computational cost.","PeriodicalId":287232,"journal":{"name":"The First Asian Conference on Pattern Recognition","volume":"139 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133912814","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}
Beyond conventional semi-supervised dimensionality reduction methods which data are represented in a single vector or graph space, multiple view semi-supervised ones are to learn a hidden consensus pattern from multiple representations of multiple view data together with some domain knowledge. Under multiple view settings, we propose a new Multiple view Semi-supervised Discriminant Analysis (MSDA). Specifically, the labeled data are used to infer the discriminant structure in each view. Simultaneously, all the data, including the labeled and the unlabeled instances, are used to discover the intrinsic geometrical structure in each view. Thus, we can learn an optimal pattern from the multiple patterns of multiple representations with serial combination after getting the projection of each view. Experiments carried out on real-world data sets by MSDA show a clear improvement over the results of representative dimensionality reduction algorithms.
{"title":"Multiple view semi-supervised discriminant analysis","authors":"Xuesong Yin, Xiaodong Chen, Xiaofang Ruan, Yarong Huang","doi":"10.1109/ACPR.2011.6166562","DOIUrl":"https://doi.org/10.1109/ACPR.2011.6166562","url":null,"abstract":"Beyond conventional semi-supervised dimensionality reduction methods which data are represented in a single vector or graph space, multiple view semi-supervised ones are to learn a hidden consensus pattern from multiple representations of multiple view data together with some domain knowledge. Under multiple view settings, we propose a new Multiple view Semi-supervised Discriminant Analysis (MSDA). Specifically, the labeled data are used to infer the discriminant structure in each view. Simultaneously, all the data, including the labeled and the unlabeled instances, are used to discover the intrinsic geometrical structure in each view. Thus, we can learn an optimal pattern from the multiple patterns of multiple representations with serial combination after getting the projection of each view. Experiments carried out on real-world data sets by MSDA show a clear improvement over the results of representative dimensionality reduction algorithms.","PeriodicalId":287232,"journal":{"name":"The First Asian Conference on Pattern Recognition","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133068153","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 : 2011-11-01DOI: 10.1109/ACPR.2011.6166544
Mingming Sun, D. Zhang, Jing-yu Yang
To improve the attractiveness of a face, intuitively, one can drive the face approaching some beautiful faces. There are two major problem to solve for implementing the intuitive solution. One problem is that how to define and discover suitable beauty prototypes. Another is that how to determine the balance between the original face and the beauty prototype to produce the desired face. In this paper, we proposed a quantitive method to solve these two problems. First, a set of beautiful face prototypes are identified as cluster centers of beautiful faces, which avoid involving specific personal facial characteristic. Second, a beauty decision function is learned as a classifier that can tell whether a face is beautiful or not. Then, the facial attractiveness improvement procedure finds the nearest beauty prototype for the original face, and then approaches the prototype from the original face until the beauty decision function tells the approaching face is beautiful. With this method, the face is beautified and the difference between the beautified face and the original face is minimized. The experimental results verify the validity of the proposed methods.
{"title":"Face attractiveness improvement using beauty prototypes and decision","authors":"Mingming Sun, D. Zhang, Jing-yu Yang","doi":"10.1109/ACPR.2011.6166544","DOIUrl":"https://doi.org/10.1109/ACPR.2011.6166544","url":null,"abstract":"To improve the attractiveness of a face, intuitively, one can drive the face approaching some beautiful faces. There are two major problem to solve for implementing the intuitive solution. One problem is that how to define and discover suitable beauty prototypes. Another is that how to determine the balance between the original face and the beauty prototype to produce the desired face. In this paper, we proposed a quantitive method to solve these two problems. First, a set of beautiful face prototypes are identified as cluster centers of beautiful faces, which avoid involving specific personal facial characteristic. Second, a beauty decision function is learned as a classifier that can tell whether a face is beautiful or not. Then, the facial attractiveness improvement procedure finds the nearest beauty prototype for the original face, and then approaches the prototype from the original face until the beauty decision function tells the approaching face is beautiful. With this method, the face is beautified and the difference between the beautified face and the original face is minimized. The experimental results verify the validity of the proposed methods.","PeriodicalId":287232,"journal":{"name":"The First Asian Conference on Pattern Recognition","volume":"42 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115576518","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 : 2011-11-01DOI: 10.1109/ACPR.2011.6166567
Zhenyu Liu, Jing Tian, Li Chen, Yongtao Wang
This paper addresses the super-resolution image reconstruction problem with the aim to produce a higher-resolution image based on its low-resolution counterparts. The proposed approach adaptively adjusts the degree of regularization using the saliency measure of the local content of the image. This is in contrast to that a spatially-invariant regularization is used for the whole image in conventional approaches. Furthermore, a gradient-based assessment criterion is proposed to measure the saliency of the image. Experiments are conducted to demonstrate the superior performance of the proposed approach.
{"title":"Spatially-adaptive regularized super-resolution image reconstruction using a gradient-based saliency measure","authors":"Zhenyu Liu, Jing Tian, Li Chen, Yongtao Wang","doi":"10.1109/ACPR.2011.6166567","DOIUrl":"https://doi.org/10.1109/ACPR.2011.6166567","url":null,"abstract":"This paper addresses the super-resolution image reconstruction problem with the aim to produce a higher-resolution image based on its low-resolution counterparts. The proposed approach adaptively adjusts the degree of regularization using the saliency measure of the local content of the image. This is in contrast to that a spatially-invariant regularization is used for the whole image in conventional approaches. Furthermore, a gradient-based assessment criterion is proposed to measure the saliency of the image. Experiments are conducted to demonstrate the superior performance of the proposed approach.","PeriodicalId":287232,"journal":{"name":"The First Asian Conference on Pattern Recognition","volume":"42 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124391036","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 : 2011-11-01DOI: 10.1109/ACPR.2011.6166670
Chenxing Wang, F. Da
To address the issue of spectrum overlapping in Fourier transform profilometry, a new method based on Bi-dimensional Empirical Mode Decomposition (BEMD) is proposed. BEMD is an adaptive data decomposition method, so it does not need filters or basic functions which are important for Fourier transform or wavelet transform. In this paper, the complicated original signal of distorted fringe pattern is decomposed into several Bi-dimensional Intrinsic Mode Functions (BIMFs) as well as the residual component, with which the background component and some other frequency noises of fringe pattern can be eliminated effectively. It is beneficial to extract the first frequency component exactly for the subsequent wrapped phase retrieval in Fourier transform. Simulation and experiments illustrate the feasibility and the exactness of the proposed method.
{"title":"A novel method of eliminating the background in Fourier transform profilometry based on Bi-dimensional Empirical Mode Decomposition","authors":"Chenxing Wang, F. Da","doi":"10.1109/ACPR.2011.6166670","DOIUrl":"https://doi.org/10.1109/ACPR.2011.6166670","url":null,"abstract":"To address the issue of spectrum overlapping in Fourier transform profilometry, a new method based on Bi-dimensional Empirical Mode Decomposition (BEMD) is proposed. BEMD is an adaptive data decomposition method, so it does not need filters or basic functions which are important for Fourier transform or wavelet transform. In this paper, the complicated original signal of distorted fringe pattern is decomposed into several Bi-dimensional Intrinsic Mode Functions (BIMFs) as well as the residual component, with which the background component and some other frequency noises of fringe pattern can be eliminated effectively. It is beneficial to extract the first frequency component exactly for the subsequent wrapped phase retrieval in Fourier transform. Simulation and experiments illustrate the feasibility and the exactness of the proposed method.","PeriodicalId":287232,"journal":{"name":"The First Asian Conference on Pattern Recognition","volume":"44 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124409902","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 : 2011-11-01DOI: 10.1109/ACPR.2011.6166575
Ruiguang Hu, Nianhua Xie, Weiming Hu
In this paper we propose an algorithm for the recognition of three kinds of drug-taking instruments, including bongs, hookahs and spoons. A global feature - Pyramid of Histograms of Orientation Gradients (PHOG) - is used to represent images. PHOG is calculated by partitioning an image into increasingly fine sub-regions and concatenating the appropriately weighted histograms of orientation gradients of each sub-region at each level. Then, different classifiers can be employed to handle this recognition problem. In our experiments, Support Vector Machines (SVM) with five different kernels and Random Forest are evaluated for our application and SVM with χ2 kernel shows the best performance. We also compare our method with the standard Bag-of-Words (BOW) model using SIFT features. Experimental results demonstrate that in our application, directly using appropriate global feature (PHOG) is better than using local feature (SIFT) and BOW model in both performance and complexity.
{"title":"Drug-taking instruments recognition","authors":"Ruiguang Hu, Nianhua Xie, Weiming Hu","doi":"10.1109/ACPR.2011.6166575","DOIUrl":"https://doi.org/10.1109/ACPR.2011.6166575","url":null,"abstract":"In this paper we propose an algorithm for the recognition of three kinds of drug-taking instruments, including bongs, hookahs and spoons. A global feature - Pyramid of Histograms of Orientation Gradients (PHOG) - is used to represent images. PHOG is calculated by partitioning an image into increasingly fine sub-regions and concatenating the appropriately weighted histograms of orientation gradients of each sub-region at each level. Then, different classifiers can be employed to handle this recognition problem. In our experiments, Support Vector Machines (SVM) with five different kernels and Random Forest are evaluated for our application and SVM with χ2 kernel shows the best performance. We also compare our method with the standard Bag-of-Words (BOW) model using SIFT features. Experimental results demonstrate that in our application, directly using appropriate global feature (PHOG) is better than using local feature (SIFT) and BOW model in both performance and complexity.","PeriodicalId":287232,"journal":{"name":"The First Asian Conference on Pattern Recognition","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124450076","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 : 2011-11-01DOI: 10.1109/ACPR.2011.6892193
Ruiling Deng, Gang Zeng, Rui Gan, H. Zha
The 3D reconstruction of buildings is a challenging research problem especially for image-based methods due to the absence of textured surfaces and difficulty in detecting high-level architectural structures. In this paper, we present an image-based reconstruction algorithm for efficiently modeling of buildings with the Manhattan-world assumption. The first key component of the algorithm is a clustering of geometric primitives (e.g. stereo points and lines) into sparse planes in Manhattan-world coordinates. The combination of such clustered planes greatly limits the possibility of building models to be reconstructed. In the second stage, we employ the graph-cut minimization to obtain an optimal model based on an energy functional that embeds image consistency, surface smoothness and Manhattanworld constraints. Real world building reconstruction results demonstrate the efficiency of the proposed algorithm in handling large scale data and its robustness against the variety of architectural structures.
{"title":"Image-based building reconstruction with Manhattan-world assumption","authors":"Ruiling Deng, Gang Zeng, Rui Gan, H. Zha","doi":"10.1109/ACPR.2011.6892193","DOIUrl":"https://doi.org/10.1109/ACPR.2011.6892193","url":null,"abstract":"The 3D reconstruction of buildings is a challenging research problem especially for image-based methods due to the absence of textured surfaces and difficulty in detecting high-level architectural structures. In this paper, we present an image-based reconstruction algorithm for efficiently modeling of buildings with the Manhattan-world assumption. The first key component of the algorithm is a clustering of geometric primitives (e.g. stereo points and lines) into sparse planes in Manhattan-world coordinates. The combination of such clustered planes greatly limits the possibility of building models to be reconstructed. In the second stage, we employ the graph-cut minimization to obtain an optimal model based on an energy functional that embeds image consistency, surface smoothness and Manhattanworld constraints. Real world building reconstruction results demonstrate the efficiency of the proposed algorithm in handling large scale data and its robustness against the variety of architectural structures.","PeriodicalId":287232,"journal":{"name":"The First Asian Conference on Pattern Recognition","volume":"42 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122026537","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：