EMPIRE10 (Evaluation of Methods for Pulmonary Image REgistration 2010) is a public platform for f... more EMPIRE10 (Evaluation of Methods for Pulmonary Image REgistration 2010) is a public platform for fair and meaningful comparison of registration algorithms which are applied to a database of intra-patient thoracic CT image pairs. Evaluation of non-rigid registration techniques is a non trivial task. This is compounded by the fact that researchers typically test only on their own data, which varies widely. For this reason, reliable assessment and comparison of different registration algorithms has been virtually impossible in the past. In this work we present the results of the launch phase of EMPIRE10 which comprised the comprehensive evaluation and comparison of 20 individual algorithms from leading academic and industrial research groups. All algorithms are applied to the same set of 30 thoracic CT pairs. Algorithm settings and parameters are chosen by researchers expert in the configuration of their own method and evaluation is independent, using the same criteria for all participants. All results are published on the EMPIRE10 website [1]. The challenge remains ongoing and open to new participants. Full results from 24 algorithms have been published [1] at the time of writing. This article details the organisation of the challenge, the data and evaluation methods and the outcome of the initial launch with 20 algorithms. The gain in knowledge and future work are discussed.
Resting-state fMRI is a promising imaging technique to evaluate functions in the human brain in h... more Resting-state fMRI is a promising imaging technique to evaluate functions in the human brain in health and disease, and it shows several practical advantages over task-based fMRI. Different hormonal stages of the female menstrual cycle and hormonal contraceptive use affect results in taskbased fMRI; it is however not yet clarified whether resting state networks are also altered. A population of 18 women with a natural cycle, and 19 women using second or third generation hormonal contraceptives was examined in a longitudinal study-design. The natural cycle group was scanned at 3 time-points (follicular phase, ovulation, luteal phase), and the contraceptives group was scanned twice (inactive pill-phase, active pill-phase). Blood samples were acquired to evaluate hormonal concentrations, and premenstrual symptoms were assessed through daily record of severity of problems questionnaires. Results show no major alterations in the default mode network and the executive control network between different hormonal phases. A positive correlation of functional connectivity in the posterior part of the default mode network (DMN) was found with premenstrual-like symptoms in the hormonal contraceptives group. Using the current methodology, the studied resting state networks seem to show a decent stability throughout menstrual cycle phases. Also, no effect of hormonal contraceptive use is found. Interestingly, we show for the first time an association of DMN alterations with periodic psychological and somatic symptoms, experienced during the inactive pill-phase by a sub-population of women.
Polyetheretherketone (PEEK) materials already have been used successfully in orthopedic and espec... more Polyetheretherketone (PEEK) materials already have been used successfully in orthopedic and especially spine surgery. PEEK is radiolucent and comparable with bone regarding elasticity. However, PEEK is inert and the adhesion of PEEK implants to bone tissue proceeds slowly because of their relatively low biocompatibility. The aim of the study is to evaluate the effect of titanium and CaP coating on the adhesion of bone tissue. Six adult sheep (body weight 57.6 ± 3.9 kg) were included in this study. Three different types of cylindrical dowels (12 mm length x 8 mm diameter) were implanted in long bones (tibia and femur): PEEK dowels without coating (the control group), and PEEK dowels with a nanocoating of calcium phosphate (CaP group) or titanium (titanium group). Animals were sacrificed after 6, 12 and 26 weeks. Dowels were explanted for micro CT and histology. Bone implant contact (BIC) ratio was significantly higher in the titanium versus control groups in the 6 to 12 weeks period (p = 0.03). The ratio between bone volume and tissue volume (BV/TV) was significantly higher in titanium versus control in the 6 to 12 weeks period (p = 0.02). A significant correlation between BIC and BV/TV was seen (r = 0.85, p < 0.05). Coating of PEEK dowels with a nanocoating of titanium has beneficial effects on adhesion of bone tissue.
In order to study ventilation or to extract other functional information of the lungs, intra-pati... more In order to study ventilation or to extract other functional information of the lungs, intra-patient matching of scans at a different in-spiration level is valuable as an examination tool. In this paper, a method for robust 3D tree matching is proposed as an independent registration method or as a guide for other, e.g. voxel-based, types of registration. For the first time, the 3D tree is represented by intrinsic matrices, reference frame independent descriptions containing the geodesic or Euclidean dis-tance between each pair of detected bifurcations. Marginalization of point pair probabilities based on the intrinsic matrices provides soft assign cor-respondences between the two trees. This global correspondence model is combined with local bifurcation similarity models, based on the shape of the adjoining vessels and the local gray value distribution. As a proof of concept of this general matching approach, the method is applied for matching lung vessel trees acquired from CT imag...
In this paper a speciflc method is presented to facilitate the semi-automatic segmentation of liv... more In this paper a speciflc method is presented to facilitate the semi-automatic segmentation of liver metastases in CT images. Accurate and reliable segmentation of tumors is e.g. essential for the follow-up of cancer treatment. The core of the algorithm is a level set function. The initialization is provided by a spiral-scanning technique based on dy- namic programming. The level set
2010 Fourth IEEE International Conference on Biometrics: Theory, Applications and Systems (BTAS), 2010
This paper presents a SIFT algorithm adapted for 3D surfaces (called meshSIFT) and its applicatio... more This paper presents a SIFT algorithm adapted for 3D surfaces (called meshSIFT) and its applications to 3D face pose normalisation and recognition. The algorithm allows reliable detection of scale space extrema as local feature locations. The scale space contains the mean curvature in each vertex on different smoothed versions of the input mesh. The meshSIFT algorithm then describes the neighbourhood of every scale space extremum in a feature vector consisting of concatenated histograms of shape indices and slant angles. The feature vectors are reliably matched by comparing the angle in feature space.
4th International Conference on Imaging for Crime Detection and Prevention 2011 (ICDP 2011), 2011
Estimating the facial outlook from an unidentified skull is a challenging task in forensic invest... more Estimating the facial outlook from an unidentified skull is a challenging task in forensic investigations. This paper presents the definition and implementation of a craniofacial model for computerized craniofacial reconstruction (CFR). The craniofacial model consists of a craniofacial template that is warped towards an unidentified target skull. The allowed transformations for this warping are statistically defined using a PCAbased transformation model, resulting in a linear combination of major modes of deformations. This work builds on previous work in which a statistical model was constructed based on facial shape (represented as a dense set of points) variations and sparse soft tissue depths at 52 craniofacial landmarks. The main contribution of this work is the extension of the soft tissue depth measurements to a dense set of points derived from a database of head CT-images of 156 patients. Despite the limited amount of training data compared to the number of degrees of freedom, the reconstruction tests show good results for a larger part of the test data. Root mean squared error (RMSE) values between reconstruction results and ground truth data smaller than 4 mm over the total head and neck region are observed.
2012 9th IEEE International Symposium on Biomedical Imaging (ISBI), 2012
Piecewise rigid registration is a fundamental problem in medical imaging involving intra-patient ... more Piecewise rigid registration is a fundamental problem in medical imaging involving intra-patient pose differences in multiple medical images, mostly due to articulated motion. In this paper, we propose a method to extract multiple rigid transformations in 2D medical images in the presence of outliers. First, points of interest in the images are extracted and matched with the SIFT algorithm. Secondly, multiple rigid motions are sampled and clustered by the mean shift algorithm in the special Euclidean group SE(2), a smooth manifold of 2-D rigid transformation matrices. The method proposed is evaluated for intra-subject registrations of knee fluoroscopy images, demonstrating a mean angular and translational error on the estimated motion of 0.39 • and 6.65 pixels, respectively.
ABSTRACT Intra-shape deformations complicate 3D shape recognition and therefore need proper model... more ABSTRACT Intra-shape deformations complicate 3D shape recognition and therefore need proper modeling. Thereto, an isometric deformation model is used in this paper. The method proposed does not need explicit point correspondences for the comparison of 3D shapes. The geodesic distance matrix is used as an isometry-invariant shape representation. Two approaches are described to arrive at a sampling order invariant shape descriptor: the histogram of geodesic distance matrix values and the set of largest singular values of the geodesic distance matrix. Shape comparison is performed by comparison of the shape descriptors using the χ2-distanceχ2-distance as dissimilarity measure. For object recognition, the results obtained demonstrate the singular value approach to outperform the histogram-based approach, as well as the state-of-the-art multidimensional scaling technique, the ICP baseline algorithm and other isometric deformation modeling methods found in literature. Using the TOSCA database, a rank-1 recognition rate of 100% is obtained for the identification scenario, while the verification experiments are characterized by a 1.58% equal error rate. External validation demonstrates that the singular value approach outperforms all other participants for the non-rigid object retrieval contests in SHREC 2010 as well as SHREC 2011. For 3D face recognition, the rank-1 recognition rate is 61.9% and the equal error rate is 11.8% on the BU-3DFE database. This decreased performance is attributed to the fact that the isometric deformation assumption only holds to a limited extent for facial expressions. This is also demonstrated in this paper.
This paper addresses the problem of establishing point correspondences between two object instanc... more This paper addresses the problem of establishing point correspondences between two object instances using spectral high-order graph matching. Therefore, 3D objects are intrinsically represented by weighted high-order adjacency tensors. These are, depending on the weighting scheme, invariant for structure-preserving, equi-areal, conformal or volume-preserving object deformations. Higher-order spectral decomposition transforms the NP-hard assignment problem into a linear assignment problem by canonical embedding. This allows to extract dense correspondence information with reasonable computational complexity, making the method faster than any other previously published method imposing higher-order constraints to shape matching. Robustness against missing data and resampling is measured and compared with a baseline spectral graph matching method.
In this paper the problem of pairwise model-to-scene point set registration is considered. Three ... more In this paper the problem of pairwise model-to-scene point set registration is considered. Three contributions are made. Firstly, the relations between correspondencebased and some information-theoretic point cloud registration algorithms are formalized. Starting from the observation that the outlier handling of existing methods relies on heuristically determined models, a second contribution is made exploiting aforementioned relations to derive a new robust point set registration algorithm. Representing model and scene point cloud by mixtures of Gaussians, the method minimizes their Kullback-Leibler divergence both w.r.t. the registration transformation parameters and w.r.t. the scene's model mixture coefficients. This results in an Expectation-Maximization Iterative Closest Point (EM-ICP) approach with a parameter-free outlier model that is optimal in information-theoretic sense. While the current (CUDA) implementation is limited to the rigid registration case, the underlying theory applies to both rigid and nonrigid point set registration. As a by-product of the registration algorithm's theory, a third contribution is made by suggesting a new point cloud Kernel Density Estimation approach which relies on maximizing the resulting distribution's entropy w.r.t. the kernel weights.
2010 20th International Conference on Pattern Recognition, 2010
The recognition of faces under varying expressions is one of the current challenges in the face r... more The recognition of faces under varying expressions is one of the current challenges in the face recognition community. In this paper, we propose a method fusing different complementary approaches each dealing with expression variations. The first approach uses an isometric deformation model and is based on the largest singular values of the geodesic distance matrix as an expression-invariant shape descriptor. The second approach performs recognition on the more rigid parts of the face that are less affected by expression variations. Several fusion techniques are examined for combining the approaches. The presented method is validated on a subset of 900 faces of the BU-3DFE face database resulting in an equal error rate of 5.85% for the verification scenario and a rank 1 recognition rate of 94.48% for the identification scenario using the sum rule as fusion technique. This result outperforms other 3D expression-invariant face recognition methods on the same database.
2009 IEEE 3rd International Conference on Biometrics: Theory, Applications, and Systems, 2009
Currently, the recognition of faces under varying expressions is one of the main challenges in th... more Currently, the recognition of faces under varying expressions is one of the main challenges in the face recognition community. In this paper a method is presented dealing with those expression variations by using an isometric deformation model. The method is built upon the geodesic distance matrix as a representation of the 3D face. We will show that the set of largest singular values is an excellent expression-invariant shape descriptor. Face comparison is performed by comparison of their shape descriptors using the mean normalized Manhattan distance as dissimilarity measure. The presented method is validated on a subset of 900 faces of the BU-3DFE face database resulting in an equal error rate of 13.37% for the verification scenario. This result is comparable with the equal error rates of other 3D expression-invariant face recognition methods using an isometric deformation model on the same database.
Eurographics Workshop on 3D Object Retrieval, EG 3DOR, 2010
SHREC'10 Track: Non-rigid 3D Shape Retrieval Lian, Z.; Godil, A.; Fabry, T.; Furuya, T.; Hermans,... more SHREC'10 Track: Non-rigid 3D Shape Retrieval Lian, Z.; Godil, A.; Fabry, T.; Furuya, T.; Hermans, J.; Ohbuchi, R.; Shu, C.; Smeets, D.; Suetens, P.; Vandermeulen, D.; Wuhrer, S. Contact us / Contactez nous: nparc.cisti@nrc-cnrc.gc.ca. http://nparc.cisti-icist.nrc-cnrc.gc.ca/npsi/jsp/nparc_cp.jsp?lang=fr L'accès à ce site Web et l'utilisation de son contenu sont assujettis aux conditions présentées dans le site
Since most 3D cameras cannot capture the complete 3D face, an important challenge in 3D face reco... more Since most 3D cameras cannot capture the complete 3D face, an important challenge in 3D face recognition is the comparison of two 3D facial surfaces with little or no overlap. In this paper, a local feature method is presented to tackle this challenge exploiting the symmetry of the human face. Features are located and described using an extension of SIFT for meshes (meshSIFT). As such, features are localized as extrema in the curvature scale space of the input mesh, and are described by concatenating histograms of shape indices and slant angles of the neighborhood. For 3D face scans with sufficient overlap, the number of matching meshSIFT features is a reliable measure for face recognition purposes. However, as the feature descriptor is not symmetrical, features on one face are not matched with their symmetrical counterpart on another face impeding their feasibility for comparison of face scans with limited or no (left-right) overlap. In order to alleviate this problem, facial symmetry could be used to increase the overlap between two face scans by mirroring one of both faces w.r.t. an arbitrary plane. As this would increase the computational demand, this paper proposes an efficient approach to describe the features of a mirrored face by mirroring the mesh-SIFT descriptors of the input face. The presented method is validated on the data of the "SHREC '11: Face Scans" contest, containing many partial scans. This resulted in a recognition rate of 98.6% and a mean average precision of 93.3%, clearly outperforming all other participants in the challenge.
We present two methods for isometrically deformable object recognition. The methods are built upo... more We present two methods for isometrically deformable object recognition. The methods are built upon the use of geodesic distance matrices (GDM) as an object representation. The first method compares these matrices by using histogram comparisons. The second method is a modal approach. The largest singular values or eigenvalues appear to be an excellent shape descriptor, based on the comparison with other methods also using the isometric deformation model and a general baseline algorithm. The methods are validated using the TOSCA database of non-rigid objects and a rank 1 recognition rate of 100% is reported for the modal representation method using the 50 largest eigenvalues. This is clearly higher than other methods using an isometric deformation model.
Intra-shape deformations complicate 3D object recognition and retrieval and need therefore proper... more Intra-shape deformations complicate 3D object recognition and retrieval and need therefore proper modeling. A method for inelastic deformation invariant object recognition is proposed, representing 3D objects by diffusion distance tensors (DDT), i.e. third order tensors containing the average diffusion distance for different diffusion times between each pair of points on the surface. In addition to the DDT, also geodesic distance matrices (GDM) are used to represent the objects independent of the reference frame. Transforming these distance tensors into modal representations provides a sampling order invariant shape descriptor. Different dissimilarity measures can be used for comparing these shape descriptors. The final object pair dissimilarity is the sum or product of the dissimilarities obtained by modal representations of the GDM and DDT. The method is validated on the TOSCA non-rigid world database and the SHREC 2010 dataset of non-rigid 3D models indicating that our method combining these two representations provides a more noise robust but still inter-subject shape variation sensitive method for the identification and the verification scenario in object retrieval.
ABSTRACT In this paper an automated method is presented for the localization of cephalometric lan... more ABSTRACT In this paper an automated method is presented for the localization of cephalometric landmarks in craniofacial cone-beam computed tomography images. This method makes use of a statistical sparse appearance and shape model obtained from training data. The sparse appearance model captures local image intensity patterns around each landmark. The sparse shape model, on the other hand, is constructed by embedding the landmarks in a graph. The edges of this graph represent pairwise spatial dependencies between landmarks, hence leading to a sparse shape model. The edges connecting different landmarks are defined in an automated way based on the intrinsic topology present in the training data. A maximum a posteriori approach is employed to obtain an energy function. To minimize this energy function, the problem is discretized by considering a finite set of candidate locations for each landmark, leading to a labeling problem. Using a leave-one-out approach on the training data the overall accuracy of the method is assessed. The mean and median error values for all landmarks are equal to 2.41mm\textrm{mm} and 1.49mm\textrm{mm}, respectively, demonstrating a clear improvement over previously published methods.
During the last decade research in face recognition has shifted from 2D to 3D face representation... more During the last decade research in face recognition has shifted from 2D to 3D face representations. The need for 3D face data has resulted in the advent of 3D databases. In this paper, we first give an overview of publicly available 3D face databases containing expression variations, since these variations are an important challenge in today's research. The existence of
EMPIRE10 (Evaluation of Methods for Pulmonary Image REgistration 2010) is a public platform for f... more EMPIRE10 (Evaluation of Methods for Pulmonary Image REgistration 2010) is a public platform for fair and meaningful comparison of registration algorithms which are applied to a database of intra-patient thoracic CT image pairs. Evaluation of non-rigid registration techniques is a non trivial task. This is compounded by the fact that researchers typically test only on their own data, which varies widely. For this reason, reliable assessment and comparison of different registration algorithms has been virtually impossible in the past. In this work we present the results of the launch phase of EMPIRE10 which comprised the comprehensive evaluation and comparison of 20 individual algorithms from leading academic and industrial research groups. All algorithms are applied to the same set of 30 thoracic CT pairs. Algorithm settings and parameters are chosen by researchers expert in the configuration of their own method and evaluation is independent, using the same criteria for all participants. All results are published on the EMPIRE10 website [1]. The challenge remains ongoing and open to new participants. Full results from 24 algorithms have been published [1] at the time of writing. This article details the organisation of the challenge, the data and evaluation methods and the outcome of the initial launch with 20 algorithms. The gain in knowledge and future work are discussed.
Resting-state fMRI is a promising imaging technique to evaluate functions in the human brain in h... more Resting-state fMRI is a promising imaging technique to evaluate functions in the human brain in health and disease, and it shows several practical advantages over task-based fMRI. Different hormonal stages of the female menstrual cycle and hormonal contraceptive use affect results in taskbased fMRI; it is however not yet clarified whether resting state networks are also altered. A population of 18 women with a natural cycle, and 19 women using second or third generation hormonal contraceptives was examined in a longitudinal study-design. The natural cycle group was scanned at 3 time-points (follicular phase, ovulation, luteal phase), and the contraceptives group was scanned twice (inactive pill-phase, active pill-phase). Blood samples were acquired to evaluate hormonal concentrations, and premenstrual symptoms were assessed through daily record of severity of problems questionnaires. Results show no major alterations in the default mode network and the executive control network between different hormonal phases. A positive correlation of functional connectivity in the posterior part of the default mode network (DMN) was found with premenstrual-like symptoms in the hormonal contraceptives group. Using the current methodology, the studied resting state networks seem to show a decent stability throughout menstrual cycle phases. Also, no effect of hormonal contraceptive use is found. Interestingly, we show for the first time an association of DMN alterations with periodic psychological and somatic symptoms, experienced during the inactive pill-phase by a sub-population of women.
Polyetheretherketone (PEEK) materials already have been used successfully in orthopedic and espec... more Polyetheretherketone (PEEK) materials already have been used successfully in orthopedic and especially spine surgery. PEEK is radiolucent and comparable with bone regarding elasticity. However, PEEK is inert and the adhesion of PEEK implants to bone tissue proceeds slowly because of their relatively low biocompatibility. The aim of the study is to evaluate the effect of titanium and CaP coating on the adhesion of bone tissue. Six adult sheep (body weight 57.6 ± 3.9 kg) were included in this study. Three different types of cylindrical dowels (12 mm length x 8 mm diameter) were implanted in long bones (tibia and femur): PEEK dowels without coating (the control group), and PEEK dowels with a nanocoating of calcium phosphate (CaP group) or titanium (titanium group). Animals were sacrificed after 6, 12 and 26 weeks. Dowels were explanted for micro CT and histology. Bone implant contact (BIC) ratio was significantly higher in the titanium versus control groups in the 6 to 12 weeks period (p = 0.03). The ratio between bone volume and tissue volume (BV/TV) was significantly higher in titanium versus control in the 6 to 12 weeks period (p = 0.02). A significant correlation between BIC and BV/TV was seen (r = 0.85, p < 0.05). Coating of PEEK dowels with a nanocoating of titanium has beneficial effects on adhesion of bone tissue.
In order to study ventilation or to extract other functional information of the lungs, intra-pati... more In order to study ventilation or to extract other functional information of the lungs, intra-patient matching of scans at a different in-spiration level is valuable as an examination tool. In this paper, a method for robust 3D tree matching is proposed as an independent registration method or as a guide for other, e.g. voxel-based, types of registration. For the first time, the 3D tree is represented by intrinsic matrices, reference frame independent descriptions containing the geodesic or Euclidean dis-tance between each pair of detected bifurcations. Marginalization of point pair probabilities based on the intrinsic matrices provides soft assign cor-respondences between the two trees. This global correspondence model is combined with local bifurcation similarity models, based on the shape of the adjoining vessels and the local gray value distribution. As a proof of concept of this general matching approach, the method is applied for matching lung vessel trees acquired from CT imag...
In this paper a speciflc method is presented to facilitate the semi-automatic segmentation of liv... more In this paper a speciflc method is presented to facilitate the semi-automatic segmentation of liver metastases in CT images. Accurate and reliable segmentation of tumors is e.g. essential for the follow-up of cancer treatment. The core of the algorithm is a level set function. The initialization is provided by a spiral-scanning technique based on dy- namic programming. The level set
2010 Fourth IEEE International Conference on Biometrics: Theory, Applications and Systems (BTAS), 2010
This paper presents a SIFT algorithm adapted for 3D surfaces (called meshSIFT) and its applicatio... more This paper presents a SIFT algorithm adapted for 3D surfaces (called meshSIFT) and its applications to 3D face pose normalisation and recognition. The algorithm allows reliable detection of scale space extrema as local feature locations. The scale space contains the mean curvature in each vertex on different smoothed versions of the input mesh. The meshSIFT algorithm then describes the neighbourhood of every scale space extremum in a feature vector consisting of concatenated histograms of shape indices and slant angles. The feature vectors are reliably matched by comparing the angle in feature space.
4th International Conference on Imaging for Crime Detection and Prevention 2011 (ICDP 2011), 2011
Estimating the facial outlook from an unidentified skull is a challenging task in forensic invest... more Estimating the facial outlook from an unidentified skull is a challenging task in forensic investigations. This paper presents the definition and implementation of a craniofacial model for computerized craniofacial reconstruction (CFR). The craniofacial model consists of a craniofacial template that is warped towards an unidentified target skull. The allowed transformations for this warping are statistically defined using a PCAbased transformation model, resulting in a linear combination of major modes of deformations. This work builds on previous work in which a statistical model was constructed based on facial shape (represented as a dense set of points) variations and sparse soft tissue depths at 52 craniofacial landmarks. The main contribution of this work is the extension of the soft tissue depth measurements to a dense set of points derived from a database of head CT-images of 156 patients. Despite the limited amount of training data compared to the number of degrees of freedom, the reconstruction tests show good results for a larger part of the test data. Root mean squared error (RMSE) values between reconstruction results and ground truth data smaller than 4 mm over the total head and neck region are observed.
2012 9th IEEE International Symposium on Biomedical Imaging (ISBI), 2012
Piecewise rigid registration is a fundamental problem in medical imaging involving intra-patient ... more Piecewise rigid registration is a fundamental problem in medical imaging involving intra-patient pose differences in multiple medical images, mostly due to articulated motion. In this paper, we propose a method to extract multiple rigid transformations in 2D medical images in the presence of outliers. First, points of interest in the images are extracted and matched with the SIFT algorithm. Secondly, multiple rigid motions are sampled and clustered by the mean shift algorithm in the special Euclidean group SE(2), a smooth manifold of 2-D rigid transformation matrices. The method proposed is evaluated for intra-subject registrations of knee fluoroscopy images, demonstrating a mean angular and translational error on the estimated motion of 0.39 • and 6.65 pixels, respectively.
ABSTRACT Intra-shape deformations complicate 3D shape recognition and therefore need proper model... more ABSTRACT Intra-shape deformations complicate 3D shape recognition and therefore need proper modeling. Thereto, an isometric deformation model is used in this paper. The method proposed does not need explicit point correspondences for the comparison of 3D shapes. The geodesic distance matrix is used as an isometry-invariant shape representation. Two approaches are described to arrive at a sampling order invariant shape descriptor: the histogram of geodesic distance matrix values and the set of largest singular values of the geodesic distance matrix. Shape comparison is performed by comparison of the shape descriptors using the χ2-distanceχ2-distance as dissimilarity measure. For object recognition, the results obtained demonstrate the singular value approach to outperform the histogram-based approach, as well as the state-of-the-art multidimensional scaling technique, the ICP baseline algorithm and other isometric deformation modeling methods found in literature. Using the TOSCA database, a rank-1 recognition rate of 100% is obtained for the identification scenario, while the verification experiments are characterized by a 1.58% equal error rate. External validation demonstrates that the singular value approach outperforms all other participants for the non-rigid object retrieval contests in SHREC 2010 as well as SHREC 2011. For 3D face recognition, the rank-1 recognition rate is 61.9% and the equal error rate is 11.8% on the BU-3DFE database. This decreased performance is attributed to the fact that the isometric deformation assumption only holds to a limited extent for facial expressions. This is also demonstrated in this paper.
This paper addresses the problem of establishing point correspondences between two object instanc... more This paper addresses the problem of establishing point correspondences between two object instances using spectral high-order graph matching. Therefore, 3D objects are intrinsically represented by weighted high-order adjacency tensors. These are, depending on the weighting scheme, invariant for structure-preserving, equi-areal, conformal or volume-preserving object deformations. Higher-order spectral decomposition transforms the NP-hard assignment problem into a linear assignment problem by canonical embedding. This allows to extract dense correspondence information with reasonable computational complexity, making the method faster than any other previously published method imposing higher-order constraints to shape matching. Robustness against missing data and resampling is measured and compared with a baseline spectral graph matching method.
In this paper the problem of pairwise model-to-scene point set registration is considered. Three ... more In this paper the problem of pairwise model-to-scene point set registration is considered. Three contributions are made. Firstly, the relations between correspondencebased and some information-theoretic point cloud registration algorithms are formalized. Starting from the observation that the outlier handling of existing methods relies on heuristically determined models, a second contribution is made exploiting aforementioned relations to derive a new robust point set registration algorithm. Representing model and scene point cloud by mixtures of Gaussians, the method minimizes their Kullback-Leibler divergence both w.r.t. the registration transformation parameters and w.r.t. the scene's model mixture coefficients. This results in an Expectation-Maximization Iterative Closest Point (EM-ICP) approach with a parameter-free outlier model that is optimal in information-theoretic sense. While the current (CUDA) implementation is limited to the rigid registration case, the underlying theory applies to both rigid and nonrigid point set registration. As a by-product of the registration algorithm's theory, a third contribution is made by suggesting a new point cloud Kernel Density Estimation approach which relies on maximizing the resulting distribution's entropy w.r.t. the kernel weights.
2010 20th International Conference on Pattern Recognition, 2010
The recognition of faces under varying expressions is one of the current challenges in the face r... more The recognition of faces under varying expressions is one of the current challenges in the face recognition community. In this paper, we propose a method fusing different complementary approaches each dealing with expression variations. The first approach uses an isometric deformation model and is based on the largest singular values of the geodesic distance matrix as an expression-invariant shape descriptor. The second approach performs recognition on the more rigid parts of the face that are less affected by expression variations. Several fusion techniques are examined for combining the approaches. The presented method is validated on a subset of 900 faces of the BU-3DFE face database resulting in an equal error rate of 5.85% for the verification scenario and a rank 1 recognition rate of 94.48% for the identification scenario using the sum rule as fusion technique. This result outperforms other 3D expression-invariant face recognition methods on the same database.
2009 IEEE 3rd International Conference on Biometrics: Theory, Applications, and Systems, 2009
Currently, the recognition of faces under varying expressions is one of the main challenges in th... more Currently, the recognition of faces under varying expressions is one of the main challenges in the face recognition community. In this paper a method is presented dealing with those expression variations by using an isometric deformation model. The method is built upon the geodesic distance matrix as a representation of the 3D face. We will show that the set of largest singular values is an excellent expression-invariant shape descriptor. Face comparison is performed by comparison of their shape descriptors using the mean normalized Manhattan distance as dissimilarity measure. The presented method is validated on a subset of 900 faces of the BU-3DFE face database resulting in an equal error rate of 13.37% for the verification scenario. This result is comparable with the equal error rates of other 3D expression-invariant face recognition methods using an isometric deformation model on the same database.
Eurographics Workshop on 3D Object Retrieval, EG 3DOR, 2010
SHREC'10 Track: Non-rigid 3D Shape Retrieval Lian, Z.; Godil, A.; Fabry, T.; Furuya, T.; Hermans,... more SHREC'10 Track: Non-rigid 3D Shape Retrieval Lian, Z.; Godil, A.; Fabry, T.; Furuya, T.; Hermans, J.; Ohbuchi, R.; Shu, C.; Smeets, D.; Suetens, P.; Vandermeulen, D.; Wuhrer, S. Contact us / Contactez nous: nparc.cisti@nrc-cnrc.gc.ca. http://nparc.cisti-icist.nrc-cnrc.gc.ca/npsi/jsp/nparc_cp.jsp?lang=fr L'accès à ce site Web et l'utilisation de son contenu sont assujettis aux conditions présentées dans le site
Since most 3D cameras cannot capture the complete 3D face, an important challenge in 3D face reco... more Since most 3D cameras cannot capture the complete 3D face, an important challenge in 3D face recognition is the comparison of two 3D facial surfaces with little or no overlap. In this paper, a local feature method is presented to tackle this challenge exploiting the symmetry of the human face. Features are located and described using an extension of SIFT for meshes (meshSIFT). As such, features are localized as extrema in the curvature scale space of the input mesh, and are described by concatenating histograms of shape indices and slant angles of the neighborhood. For 3D face scans with sufficient overlap, the number of matching meshSIFT features is a reliable measure for face recognition purposes. However, as the feature descriptor is not symmetrical, features on one face are not matched with their symmetrical counterpart on another face impeding their feasibility for comparison of face scans with limited or no (left-right) overlap. In order to alleviate this problem, facial symmetry could be used to increase the overlap between two face scans by mirroring one of both faces w.r.t. an arbitrary plane. As this would increase the computational demand, this paper proposes an efficient approach to describe the features of a mirrored face by mirroring the mesh-SIFT descriptors of the input face. The presented method is validated on the data of the "SHREC '11: Face Scans" contest, containing many partial scans. This resulted in a recognition rate of 98.6% and a mean average precision of 93.3%, clearly outperforming all other participants in the challenge.
We present two methods for isometrically deformable object recognition. The methods are built upo... more We present two methods for isometrically deformable object recognition. The methods are built upon the use of geodesic distance matrices (GDM) as an object representation. The first method compares these matrices by using histogram comparisons. The second method is a modal approach. The largest singular values or eigenvalues appear to be an excellent shape descriptor, based on the comparison with other methods also using the isometric deformation model and a general baseline algorithm. The methods are validated using the TOSCA database of non-rigid objects and a rank 1 recognition rate of 100% is reported for the modal representation method using the 50 largest eigenvalues. This is clearly higher than other methods using an isometric deformation model.
Intra-shape deformations complicate 3D object recognition and retrieval and need therefore proper... more Intra-shape deformations complicate 3D object recognition and retrieval and need therefore proper modeling. A method for inelastic deformation invariant object recognition is proposed, representing 3D objects by diffusion distance tensors (DDT), i.e. third order tensors containing the average diffusion distance for different diffusion times between each pair of points on the surface. In addition to the DDT, also geodesic distance matrices (GDM) are used to represent the objects independent of the reference frame. Transforming these distance tensors into modal representations provides a sampling order invariant shape descriptor. Different dissimilarity measures can be used for comparing these shape descriptors. The final object pair dissimilarity is the sum or product of the dissimilarities obtained by modal representations of the GDM and DDT. The method is validated on the TOSCA non-rigid world database and the SHREC 2010 dataset of non-rigid 3D models indicating that our method combining these two representations provides a more noise robust but still inter-subject shape variation sensitive method for the identification and the verification scenario in object retrieval.
ABSTRACT In this paper an automated method is presented for the localization of cephalometric lan... more ABSTRACT In this paper an automated method is presented for the localization of cephalometric landmarks in craniofacial cone-beam computed tomography images. This method makes use of a statistical sparse appearance and shape model obtained from training data. The sparse appearance model captures local image intensity patterns around each landmark. The sparse shape model, on the other hand, is constructed by embedding the landmarks in a graph. The edges of this graph represent pairwise spatial dependencies between landmarks, hence leading to a sparse shape model. The edges connecting different landmarks are defined in an automated way based on the intrinsic topology present in the training data. A maximum a posteriori approach is employed to obtain an energy function. To minimize this energy function, the problem is discretized by considering a finite set of candidate locations for each landmark, leading to a labeling problem. Using a leave-one-out approach on the training data the overall accuracy of the method is assessed. The mean and median error values for all landmarks are equal to 2.41mm\textrm{mm} and 1.49mm\textrm{mm}, respectively, demonstrating a clear improvement over previously published methods.
During the last decade research in face recognition has shifted from 2D to 3D face representation... more During the last decade research in face recognition has shifted from 2D to 3D face representations. The need for 3D face data has resulted in the advent of 3D databases. In this paper, we first give an overview of publicly available 3D face databases containing expression variations, since these variations are an important challenge in today's research. The existence of
Uploads
Papers by Dirk Smeets