Thesis (Ph. D.)--University of Hong Kong, 2006. / Title proper from title frame. Also available in printed format.

Thesis (Ph. D.)--University of Hong Kong, 2006. / Title proper from title frame. Also available in printed format.

La reconstruction faciale à partir d'un squelette crânien est une technique importante dans plusieurs domaines scientifiques, en particulier dans les sciences médico-légales, l'archéologie et la paléontologie pour l'identification de crânes et la reconnaissance de personnes. Elle fait partie des méthodes d'identification reconstructive et est utilisée le plus souvent en dernier recours, lorsqu'aucune autre technique ne permet de présumer l'identité inconnue de la personne. Cette thèse aborde la problématique de la reconstruction faciale à partir d'images tridimensionnelles (3D) de crânes humains considérés comme étant inconnus. Nous présentons une méthode de reconstruction faciale 3D informatisée reposant sur des techniques récentes d'imagerie médicale avec comme principaux objectifs la rapidité du traitement et l'élimination de la subjectivité en s'appuyant en particulier sur des critères mathématiques pour évaluer les résultats. Notre méthode est fondée sur une approche par modèle déformable contraint par la connaissance des épaisseurs des tissus mous en un certain nombre de points de repère caractéristiques. Elle utilise, pour chaque reconstruction faciale, une image 3D d'une tête de référence dont nous extrayons la peau et le crâne, et une image 3D du crâne d'une tête inconnue dont nous voulons reconstruire la peau. La procédure de la reconstruction faciale est divisée en deux principales étapes. Une étape d'initialisation du modèle déformable est fondée sur une technique de recalage non linéaire guidé par un modèle de déformations de forme libre (FFD) à base de B-splines. Nous proposons donc une initialisation automatique, réalisée uniquement à partir d'un ensemble crâne et peau de référence et du crâne inconnu. Nous calculons, dans un premier temps, une transformation de l'image 3D d'un crâne de référence vers celle du crâne inconnu. Ensuite, nous appliquons la même transformation pour déformer l'image de la peau de référence vers une nouvelle peau que nous considérons proche de la peau inconnue et qui servira d'initialisation à la reconstruction faciale finale. Dans une seconde étape, la peau initiale, résultant de la transformation calculée précédemment, est raffinée à l'aide d'un modèle déformable 3D à base de maillages simplexes qui est attiré par un ensemble des points de repère caractéristiques préalablement calculés par un calcul des courbures moyenne et gaussienne, et fixés selon les positions des repères anthropologiques de Rhine et Campbell [Rhine et Campbell, 1980]. L'évolution de notre modèle déformable est effectuée suivant différentes forces internes et externes dont la force de champ de vecteurs de gradients (GVF) et une force de pression. Cette méthode a été testée sur treize ensembles de données crâne/peau issus d'IRM-3D de têtes d'individus enfants et adultes. Pour chaque reconstruction faciale, deux ensembles crâne/peau correspondant à une tête de référence et à une tête dont la peau est à reconstruire sont sélectionnés selon des caractéristiques anthropologiques similaires. Par cette méthode, nous obtenons des résultats encourageants. Les formes reconstruites restent des visages, visiblement acceptables, et sont relativement proches des visages réels.

<p>Methods for measuring deformation and motion of the human heart in-vivo are crucial in the assessment of cardiac function. Applications ranging from basic physiological research, through early detection of disease to follow-up studies, all benefit from improved methods of measuring the dynamics of the heart. This thesis presents new methods for acquisition, reconstruction and visualization of cardiac motion and deformation, based on magnetic resonance imaging.</p><p>Local heart wall deformation can be quantified in a strain rate tensor field. This tensor field describes the local deformation excluding rigid body translation and rotation. The drawback of studying this tensor-valued quantity, as opposed to a velocity vector field, is the high dimensionality of the tensor. The problem of visualizing the tensor field is approached by combining a local visualization that displays all degrees of freedom for a single tensor with an overview visualization using a scalar field representation of the complete tensor field. The scalar field is obtained by iterated adaptive filtering of a noise field.</p><p>Several methods for synchronizing the magnetic resonance imaging acquisition to the heart beat have previously been used to resolve individual heart phases from multiple cardiac cycles. In the present work, one of these techniques is extended to resolve two temporal dimensions simultaneously, the cardiac cycle and the respiratory cycle. This is combined with volumetric imaging to produce a five-dimensional data set. Furthermore, the acquisition order is optimized in order to reduce eddy current artifacts.</p><p>The five-dimensional acquisition either requires very long scan times or can only provide low spatiotemporal resolution. A method that exploits the variation in temporal bandwidth over the imaging volume, k-t BLAST, is described and extended to two simultaneous temporal dimensions. The new method, k-t2 BLAST, allows simultaneous reduction of scan time and improvement of spatial resolution.</p> / Report code: LIU-TEK-LIC-2006:43

In AR systems, registration is one of the most difficult problems currently limiting their applications. In this paper, we proposed a simple registration method using projective reconstruction. This method consists of two steps: embedding and tracking. Embedding involves specifying four points to build the world coordinate system on which a virtual object will be superimposed. In tracking, a projective reconstruction technique in computer vision is used to track the four specified points to compute the modelview transformation for augmentation. This method is simple as only four points need to be specified at the embedding stage, and the virtual object can then be easily augmented in a real video sequence. In addition, it can be extended to a common scenario using a common projective matrix. The proposed method has three advantages: (1) It is fast because the linear least square method can be used to estimate the related matrix in the algorithm and it is not necessary to calculate the fundamental matrix in the extended case; (2) A virtual object can still be superimposed on a related area even if some parts of the specified area are occluded during the augmentation process; and (3) This method is robust because it remains effective even when not all the reference points are detected during the augmentation process (in the rendering process), as long as at least six pairs of related reference point correspondences can be found. Several projective matrices obtained from the authors’ previous work, which are unrelated with the present AR system, were tested on this extended registration method. Experiments showed that these projective matrices can also be utilized for tracking the specified points. / Singapore-MIT Alliance (SMA)

This paper presents the theory behind a model for a two-stage analog network for edge detection and image reconstruction to be implemented in VLSI. Edges are detected in the first stage using the multi-scale veto rule, which eliminates candidates that do not pass a threshold test at each of a set of different spatial scales. The image is reconstructed in the second stage from the brightness values adjacent to edge locations. The MSV rule allows good localization and efficient noise removal. Since the reconstructed images are visually similar to the originals, the possibility exists of achieving significant bandwidth compression.

Computed tomography imaging spectrometer (CTIS) technology is introduced and its use is discussed. An iterative method is presented for CTIS image-reconstruction in the presence of both photon noise in the image and post-detection Gaussian system noise. The new algorithm assumes the transfer matrix of the system has a particular structure. Error analysis, performance evaluation, and parallelization of the algorithm is done. Complexity analysis is performed for the proof of concept code developed. Future work is discussed relating to potential improvements to the algorithm. An intuitive explanation for the success of the new algorithm is that it reformulates the image reconstruction problem as a constrained problem such that an explicit closed form solution can be computed when the constraint is ignored. Incorporating the constraint leads to an inverse matrix problem which can be dealt with using a conjugate gradient method. A weighted iterative refinement technique is employed because the conjugate gradient solver is terminated prematurely. This dissertation makes the following contributions to the state of the art. First, our method is several orders of magnitude faster that the previous industry best (multiplicative algebraic reconstruction technique (MART) and mixed-expectation reconstruction technique (MERT)). Second, error bounds are established. Third, open source proof of concept code is made available.

Multiple reconstructions of April 1st snow water equivalent (SWE) are generated for the Wind River Range (WRR), located in west-central Wyoming, to determine the most accurate predictors. Predictors included climate signal data (Southern Oscillation Index), traditional predictors (tree-ring chronologies), and non-spatially biased Pacific Ocean sea surface temperatures (SSTs). Incorporation of Pacific Ocean SSTs as a whole provides a more comprehensive representation of oceanic-atmospheric variability. Rotated principal component analysis (PCA) was used to regionalize April 1st snowpack data (1961 – 1999) from snow telemetry stations (SNOTEL stations). Tree-ring chronologies that were stable across the period of overlapping records (1961 – 1999) and that were positively correlated with regional snowpack at 99% confidence levels or higher were retained. Singular value decomposition (SVD) was performed on Pacific Ocean SSTs and regional snowpack data to identify coupled regions of climate (SSTs) and hydrology (SWE). Stepwise regressions were performed across the calibration period to identify the best predictor combinations. When data from the instrumental based SST regions identified by SVD were included in the pool of predictors, an increase in reconstruction skill was observed. Further regressions were performed using tree based and coral based SST data. Reconstruction equations were obtained from these regressions and regional April 1st snowpack was reconstructed for the WRR for all three types of SST data. A higher degree of snowpack variance is explained by reconstructions utilizing tree based, coral based, and instrumental based data for the Pacific Ocean SST region identified by SVD than is possible utilizing only tree-ring and SOI data, indicating that non-spatially biased SSTs are excellent predictors for snowpack reconstruction in the WRR.

Thesis (PH. D.)--Columbia University. / Vita. Also available in digital form on the Internet Archive Web site.

Thesis (Ph. D.)--University of Hong Kong, 2006. / Title proper from title frame. Also available in printed format.