M. Bicego, A. A. Salah, E. Grosso, M. Tistarelli, L. Akarun
The distinction between holistic and analytical (or feature-based) approaches to face recognition is widely held to be an important dimension of face recognition research. Holistic techniques analyze the whole face in order to recognize a subject, whereas analytical methodologies are devoted to the processing of different local parts of the face. This paper proposes a principled experimental comparison between these two approaches. Local and global face processing architectures that have access to similar feature representations and classifiers are implemented and tested under the same training and testing conditions. The analysis is performed with a recognition scenario on the difficult BANCA dataset, containing images acquired in degraded and adverse conditions. Different classifiers of increasing complexity are used in each scenario, and different classifier fusion methods are used for combining the local classifiers. Our results show that holistic approaches perform accurately only with complex classifiers, whereas feature-based approaches work better with simple classifiers. We were able to show a clear boosting effect by fusing a large number of simple classifiers.
{"title":"Generalization in Holistic versus Analytic Processing of Faces","authors":"M. Bicego, A. A. Salah, E. Grosso, M. Tistarelli, L. Akarun","doi":"10.1109/ICIAP.2007.73","DOIUrl":"https://doi.org/10.1109/ICIAP.2007.73","url":null,"abstract":"The distinction between holistic and analytical (or feature-based) approaches to face recognition is widely held to be an important dimension of face recognition research. Holistic techniques analyze the whole face in order to recognize a subject, whereas analytical methodologies are devoted to the processing of different local parts of the face. This paper proposes a principled experimental comparison between these two approaches. Local and global face processing architectures that have access to similar feature representations and classifiers are implemented and tested under the same training and testing conditions. The analysis is performed with a recognition scenario on the difficult BANCA dataset, containing images acquired in degraded and adverse conditions. Different classifiers of increasing complexity are used in each scenario, and different classifier fusion methods are used for combining the local classifiers. Our results show that holistic approaches perform accurately only with complex classifiers, whereas feature-based approaches work better with simple classifiers. We were able to show a clear boosting effect by fusing a large number of simple classifiers.","PeriodicalId":118466,"journal":{"name":"14th International Conference on Image Analysis and Processing (ICIAP 2007)","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2007-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128786169","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}
This paper shows how the relaxation labelling problem can be formulated as a diffusion process on a support graph using the Fokker-Planck equation. We abstract the labelling problem using a support graph with each graph node representing a possible object-label assignment and the edge weights representing label compatibilities. Initial object-label probabilities are updated using a relaxation-like process. The update equation is the solution of the Fokker-Planck equation, and is governed by an infinitesimal generator matrix computed from the edge-weights of the support graph. Iterative updating of the label probabilities can be effected using the eigenvalues and eigenvectors of the generator matrix. We illustrate the newly developed relaxation process for the applications of data classification.
{"title":"Kernelised Relaxation Labelling using Fokker-Planck Diffusion","authors":"Hongfang Wang, E. Hancock","doi":"10.1109/ICIAP.2007.84","DOIUrl":"https://doi.org/10.1109/ICIAP.2007.84","url":null,"abstract":"This paper shows how the relaxation labelling problem can be formulated as a diffusion process on a support graph using the Fokker-Planck equation. We abstract the labelling problem using a support graph with each graph node representing a possible object-label assignment and the edge weights representing label compatibilities. Initial object-label probabilities are updated using a relaxation-like process. The update equation is the solution of the Fokker-Planck equation, and is governed by an infinitesimal generator matrix computed from the edge-weights of the support graph. Iterative updating of the label probabilities can be effected using the eigenvalues and eigenvectors of the generator matrix. We illustrate the newly developed relaxation process for the applications of data classification.","PeriodicalId":118466,"journal":{"name":"14th International Conference on Image Analysis and Processing (ICIAP 2007)","volume":"58 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2007-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129116119","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}
In this paper we show how statistical constraints can be incorporated into the surface integration process. This problem aims to reconstruct the surface height function from a noisy field of surface normals. We propose two methods that employ a statistical model that captures variations in surface height. The first uses a coupled model that captures the variation in a training set efface surfaces in both the surface normal and surface height domain. The second is based on finding the parameters of a surface height model directly from afield of surface normals. We present experiments on ground truth face data and compare the results of the two methods with an existing surface integration technique.
{"title":"Surface Integration: Two Statistical Approaches","authors":"W. Smith, E. Hancock","doi":"10.1109/ICIAP.2007.122","DOIUrl":"https://doi.org/10.1109/ICIAP.2007.122","url":null,"abstract":"In this paper we show how statistical constraints can be incorporated into the surface integration process. This problem aims to reconstruct the surface height function from a noisy field of surface normals. We propose two methods that employ a statistical model that captures variations in surface height. The first uses a coupled model that captures the variation in a training set efface surfaces in both the surface normal and surface height domain. The second is based on finding the parameters of a surface height model directly from afield of surface normals. We present experiments on ground truth face data and compare the results of the two methods with an existing surface integration technique.","PeriodicalId":118466,"journal":{"name":"14th International Conference on Image Analysis and Processing (ICIAP 2007)","volume":"59 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2007-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129128953","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}
In autoimmune diseases, HEp-2 cells are used to detect antinuclear autoantibodies through indirect immunofluorescence (IIF) method. These cells can reveal different staining patterns that are relevant to diagnostic purposes. To classify them highly specialized personnel are required, who are not always available. In this respect, a medical demand is the development of a recognition system supporting such an activity. In this paper we present a hybrid multi-expert systems (MES) based on the reduction of the multiclass learning task to several binary problems. The combination scheme, based on both classifier fusion and selection, employs reliability estimators that aim at improving the accuracy of final classification. The performance of such a hybrid system has been compared with those of a MES based only on classifier selection, showing that the hybrid approach benefits of advantages of both combination rules.
{"title":"A Hybrid Multi-Expert Systems for HEp-2 Staining Pattern Classification","authors":"P. Soda, G. Iannello","doi":"10.1109/ICIAP.2007.9","DOIUrl":"https://doi.org/10.1109/ICIAP.2007.9","url":null,"abstract":"In autoimmune diseases, HEp-2 cells are used to detect antinuclear autoantibodies through indirect immunofluorescence (IIF) method. These cells can reveal different staining patterns that are relevant to diagnostic purposes. To classify them highly specialized personnel are required, who are not always available. In this respect, a medical demand is the development of a recognition system supporting such an activity. In this paper we present a hybrid multi-expert systems (MES) based on the reduction of the multiclass learning task to several binary problems. The combination scheme, based on both classifier fusion and selection, employs reliability estimators that aim at improving the accuracy of final classification. The performance of such a hybrid system has been compared with those of a MES based only on classifier selection, showing that the hybrid approach benefits of advantages of both combination rules.","PeriodicalId":118466,"journal":{"name":"14th International Conference on Image Analysis and Processing (ICIAP 2007)","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2007-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115287923","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}
We propose a method for reconstructing 3D face shape from a camera, which captures the object face from various viewing angles. In this method, we do not directly reconstruct the shape, but estimate a small number of parameters which represent the face shape. The parameter space is constructed with principal component analysis of database of a large number of face shapes collected for different people. By the PCA, the parameter space can represent the shape difference for the faces of various persons. From the input image sequence that is captured by the moving camera, the parameters of the object face can be estimated based on optimization framework. The experiments based on the proposed method demonstrate that the proposed method can reconstruct the facial shape with the accuracy of 2.5 mm averaged error.
{"title":"Face Shape Reconstruction from Image Sequence Taken with Monocular Camera using Shape Database","authors":"H. Saito, Yosuke Ito, M. Mochimaru","doi":"10.1109/ICIAP.2007.62","DOIUrl":"https://doi.org/10.1109/ICIAP.2007.62","url":null,"abstract":"We propose a method for reconstructing 3D face shape from a camera, which captures the object face from various viewing angles. In this method, we do not directly reconstruct the shape, but estimate a small number of parameters which represent the face shape. The parameter space is constructed with principal component analysis of database of a large number of face shapes collected for different people. By the PCA, the parameter space can represent the shape difference for the faces of various persons. From the input image sequence that is captured by the moving camera, the parameters of the object face can be estimated based on optimization framework. The experiments based on the proposed method demonstrate that the proposed method can reconstruct the facial shape with the accuracy of 2.5 mm averaged error.","PeriodicalId":118466,"journal":{"name":"14th International Conference on Image Analysis and Processing (ICIAP 2007)","volume":"5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2007-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114601476","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}
In this paper we present novel techniques to adapt conventional wavelet transforms to follow locally oriented features found in images. We introduce a shift operator before each step in a lifting implementation of the DWT. The best shifts are estimated by minimizing the high-pass coefficient energy and then used in both the prediction and update lifting steps. To approximate the asymptotically optimal rate-distortion performance of a piece-wise regular function more closely, we adopt a packet wavelet decomposition. Experimental results obtained integrating the proposed transform into the JPEG2000 codec show improvements in both visual and objective tests, allowing for a "better" representation of the edges at very-low rates. Very recently, some related ideas have been presented by other authors. The most distinctive features of this paper include a more flexible packet wavelet decomposition structure and a comparison between subband- and image-domain shifting operators.
{"title":"Orientation Adaptive Discrete Packet Wavelet Decomposition via Shifting Operators for Image Compression","authors":"S. Andriani, D. Taubman","doi":"10.1109/ICIAP.2007.99","DOIUrl":"https://doi.org/10.1109/ICIAP.2007.99","url":null,"abstract":"In this paper we present novel techniques to adapt conventional wavelet transforms to follow locally oriented features found in images. We introduce a shift operator before each step in a lifting implementation of the DWT. The best shifts are estimated by minimizing the high-pass coefficient energy and then used in both the prediction and update lifting steps. To approximate the asymptotically optimal rate-distortion performance of a piece-wise regular function more closely, we adopt a packet wavelet decomposition. Experimental results obtained integrating the proposed transform into the JPEG2000 codec show improvements in both visual and objective tests, allowing for a \"better\" representation of the edges at very-low rates. Very recently, some related ideas have been presented by other authors. The most distinctive features of this paper include a more flexible packet wavelet decomposition structure and a comparison between subband- and image-domain shifting operators.","PeriodicalId":118466,"journal":{"name":"14th International Conference on Image Analysis and Processing (ICIAP 2007)","volume":"42 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2007-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126960215","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}
This paper presents a method to identify frames with significant segmentation errors in an individual's track by analysing the changes in appearance and size features along the frame sequence. The features used and compared include global colour histograms, local histograms and the bounding box' size. Experiments were carried out on 26 tracks from 4 different people across two cameras with differing illumination conditions. By fusing two local colour features with a global colour feature, probabilities of segmentation error detection as high as 83 percent of human expert-identified major segmentation errors are achieved with false alarm rates of only 3 percent. This indicates that the analysis of such features along a track can be useful in the automatic detection of significant segmentation errors. This can improve the final results of many applications that wish to use robust segmentation results from a tracked person.
{"title":"Detecting Major Segmentation Errors for a Tracked Person Using Colour Feature Analysis","authors":"Christopher S. Madden, M. Piccardi","doi":"10.1109/ICIAP.2007.51","DOIUrl":"https://doi.org/10.1109/ICIAP.2007.51","url":null,"abstract":"This paper presents a method to identify frames with significant segmentation errors in an individual's track by analysing the changes in appearance and size features along the frame sequence. The features used and compared include global colour histograms, local histograms and the bounding box' size. Experiments were carried out on 26 tracks from 4 different people across two cameras with differing illumination conditions. By fusing two local colour features with a global colour feature, probabilities of segmentation error detection as high as 83 percent of human expert-identified major segmentation errors are achieved with false alarm rates of only 3 percent. This indicates that the analysis of such features along a track can be useful in the automatic detection of significant segmentation errors. This can improve the final results of many applications that wish to use robust segmentation results from a tracked person.","PeriodicalId":118466,"journal":{"name":"14th International Conference on Image Analysis and Processing (ICIAP 2007)","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2007-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128183082","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}
D. V. Deursen, W. D. Neve, D. D. Schrijver, R. Walle
In order to create a media resource adaptation engine that supports current and future coding formats, a generic (i.e., format-agnostic) solution is needed. One way to realize this is to rely on automatically created textual descriptions of the high-level syntax of binary media resources, which can be used to steer the adaptation of media resources. MPEG-21 generic bitstream syntax schema (gBS Schema) is a tool that enables the use of format-agnostic textual descriptions or generic bitstream syntax descriptions (gBSDs). This paper introduces gBFlavor, a novel solution for the automatic generation of gBSDs. It offers the possibility to automatically generate a format-specific parser that is able to produce a gBSD for a given media resource. In this paper, the use of gBFlavor is demonstrated for the generation of gBSDs for H.264/AVC SVC compliant bitstreams. Performance results show that the proposed solution outperforms existing techniques in terms of execution speed.
{"title":"Automatic generation of generic Bitstream Syntax Descriptions applied to H.264/AVC SVC encoded video streams","authors":"D. V. Deursen, W. D. Neve, D. D. Schrijver, R. Walle","doi":"10.1109/ICIAP.2007.32","DOIUrl":"https://doi.org/10.1109/ICIAP.2007.32","url":null,"abstract":"In order to create a media resource adaptation engine that supports current and future coding formats, a generic (i.e., format-agnostic) solution is needed. One way to realize this is to rely on automatically created textual descriptions of the high-level syntax of binary media resources, which can be used to steer the adaptation of media resources. MPEG-21 generic bitstream syntax schema (gBS Schema) is a tool that enables the use of format-agnostic textual descriptions or generic bitstream syntax descriptions (gBSDs). This paper introduces gBFlavor, a novel solution for the automatic generation of gBSDs. It offers the possibility to automatically generate a format-specific parser that is able to produce a gBSD for a given media resource. In this paper, the use of gBFlavor is demonstrated for the generation of gBSDs for H.264/AVC SVC compliant bitstreams. Performance results show that the proposed solution outperforms existing techniques in terms of execution speed.","PeriodicalId":118466,"journal":{"name":"14th International Conference on Image Analysis and Processing (ICIAP 2007)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2007-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125209360","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}
S. Chambon, Antonio Moreno-Ingelmo, A. Santhanam, J. Rolland, E. Angelini, I. Bloch
This paper deals with the problem of non-linear landmark-based registration of CT (at two different instants of the breathing cycle, intermediate expirations) and PET images of thoracic regions. We propose a general method to introduce a breathing model in a registration procedure in order to simulate the instant in the breathing cycle most similar to the PET image and guarantee physiologically plausible deformations. Initial results are very promising and demonstrate the interest of this method to improve the combination of anatomical and functional images for diagnosis and oncology applications.
{"title":"CT-PET Landmark-based Lung Registration Using a Dynamic Breathing Model","authors":"S. Chambon, Antonio Moreno-Ingelmo, A. Santhanam, J. Rolland, E. Angelini, I. Bloch","doi":"10.1109/ICIAP.2007.49","DOIUrl":"https://doi.org/10.1109/ICIAP.2007.49","url":null,"abstract":"This paper deals with the problem of non-linear landmark-based registration of CT (at two different instants of the breathing cycle, intermediate expirations) and PET images of thoracic regions. We propose a general method to introduce a breathing model in a registration procedure in order to simulate the instant in the breathing cycle most similar to the PET image and guarantee physiologically plausible deformations. Initial results are very promising and demonstrate the interest of this method to improve the combination of anatomical and functional images for diagnosis and oncology applications.","PeriodicalId":118466,"journal":{"name":"14th International Conference on Image Analysis and Processing (ICIAP 2007)","volume":"83 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2007-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128317457","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}
3D shape reconstruction from 2D images is an inverse problem, and is therefore mathematically ill- posed. One solution to 3D shape reconstruction problem is to use a model based approach. This paper presents an analysis by synthesis method for solving 3D face reconstruction problems using anatomical landmarks and intensity from 2D frontal face images. To improve the quality of 3D shape reconstruction we incorporate a number of steps in analysis by synthesis framework. Firstly, we approach the 3D model construction problem by using rigid and non rigid surface registration. Secondly, we simplify the shape estimation by using multidimensional amoeba optimization to optimize shape parameters while mapping texture directly using 3D-2D alignment. Thirdly, we evaluate the quality of the 3D shape reconstruction in the context of 3D shape error as well as by visual analysis.
{"title":"Analysis of 3D Face Reconstruction","authors":"S. Amin, D. Gillies","doi":"10.1109/ICIAP.2007.25","DOIUrl":"https://doi.org/10.1109/ICIAP.2007.25","url":null,"abstract":"3D shape reconstruction from 2D images is an inverse problem, and is therefore mathematically ill- posed. One solution to 3D shape reconstruction problem is to use a model based approach. This paper presents an analysis by synthesis method for solving 3D face reconstruction problems using anatomical landmarks and intensity from 2D frontal face images. To improve the quality of 3D shape reconstruction we incorporate a number of steps in analysis by synthesis framework. Firstly, we approach the 3D model construction problem by using rigid and non rigid surface registration. Secondly, we simplify the shape estimation by using multidimensional amoeba optimization to optimize shape parameters while mapping texture directly using 3D-2D alignment. Thirdly, we evaluate the quality of the 3D shape reconstruction in the context of 3D shape error as well as by visual analysis.","PeriodicalId":118466,"journal":{"name":"14th International Conference on Image Analysis and Processing (ICIAP 2007)","volume":"67 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2007-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128858714","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：