In this thesis, we study the concept of scalable video coding as implemented in the extension to the H.264 video coding standard. Specifically, for the spatial and quality scalability scenarios, we propose an optimization algorithm based on the Soft Decision Quantization (SDQ) concept, which aims at jointly optimizing all layers being encoded. The performance of the algorithm was assessed in an SVC implementation. Experimental results show, that the proposed method significantly improves the coding efficiency when compared to an unmodified SVC encoder.

Most of the conventional communication systems use channel interleaving as well as hard decision decoding in their designs, which lead to discarding channel memory and soft-decision information. This simplification is usually done since the complexity of handling the memory or soft-decision information is rather high. In this work, we design two lossy joint source-channel coding (JSCC) schemes that do not use explicit algebraic channel coding for a recently introduced channel model, in order to take advantage of both channel memory and soft-decision information. The channel model, called the non-binary noise discrete channel with queue based noise (NBNDC-QB), obtains closed form expressions for the channel transition distribution, correlation coefficient, and many other channel properties. The channel has binary input and $2^q$-ary output and the noise is a $2^q$-ary Markovian stationary ergodic process, based on a finite queue, where $q$ is the output's soft-decision resolution. We also numerically show that the NBNDC-QB model can effectively approximate correlated Rayleigh fading channels without losing its analytical tractability. The first JSCC scheme is the so called channel optimized vector quantizer (COVQ) and the second scheme consists of a scalar quantizer, a proper index assignment, and a sequence maximum a posteriori (MAP) decoder, designed to harness the redundancy left in the quantizer's indices, the channel's soft-decision output, and noise time correlation. We also find necessary and sufficient condition when the sequence MAP decoder is reduced to an instantaneous symbol-by-symbol decoder, i.e., a simple instantaneous mapping. / Thesis (Master, Electrical & Computer Engineering) -- Queen's University, 2011-09-25 19:43:28.785

Binary Sensor Networks are widely used in target tracking and target parameter estimation. It is more computationally and financially efficient than surveillance camera systems. According to the sensing area, binary sensors are divided into disk shaped sensors and line segmented sensors. Different mathematical methods of target trajectory estimation and characterization are applied. In this thesis, we present a mathematical model of target tracking including parameter estimation (size, intrusion velocity, trajectory, etc.) with line segmented sensor networks. Software simulation and hardware experiments are built based on the model. And we further analyze how the quantization noise affects the results.

Capsule Endoscopy (CE) is a unique technique for facilitating non-invasive and practical visualization of the entire small intestine. It has attracted a critical mass of studies for improvements. Among numerous studies being performed in capsule endoscopy, tremendous efforts are being made in the development of software algorithms to identify clinically important frames in CE videos. This thesis presents a computer-assisted method which performs automated detection of CE video-frames that contain bleeding. Specifically, a methodology is proposed to classify the frames of CE videos into bleeding and non-bleeding frames. It is a Support Vector Machine (SVM) based supervised method which classifies the frames on the basis of color features derived from image-regions. Image-regions are characterized on the basis of statistical features. With 15 available candidate features, an exhaustive feature-selection is followed to obtain the best feature subset. The best feature-subset is the combination of features that has the highest bleeding discrimination ability as determined by the three performance-metrics: accuracy, sensitivity and specificity. Also, a ground truth label annotation method is proposed in order to partially automate delineation of bleeding regions for training of the classifier. The method produced promising results with sensitivity and specificity values up to 94%. All the experiments were performed separately for RGB and HSV color spaces. Experimental results show the combination of the mean planes in red and green planes to be the best feature-subset in RGB (Red-Green-Blue) color space and the combination of the mean values of all three planes of the color space to be the best feature-subset in HSV (Hue-Saturation-Value).

Capsule Endoscopy (CE) is a unique technique for facilitating non-invasive and practical visualization of the entire small intestine. It has attracted a critical mass of studies for improvements. Among numerous studies being performed in capsule endoscopy, tremendous efforts are being made in the development of software algorithms to identify clinically important frames in CE videos. This thesis presents a computer-assisted method which performs automated detection of CE video-frames that contain bleeding. Specifically, a methodology is proposed to classify the frames of CE videos into bleeding and non-bleeding frames. It is a Support Vector Machine (SVM) based supervised method which classifies the frames on the basis of color features derived from image-regions. Image-regions are characterized on the basis of statistical features. With 15 available candidate features, an exhaustive feature-selection is followed to obtain the best feature subset. The best feature-subset is the combination of features that has the highest bleeding discrimination ability as determined by the three performance-metrics: accuracy, sensitivity and specificity. Also, a ground truth label annotation method is proposed in order to partially automate delineation of bleeding regions for training of the classifier. The method produced promising results with sensitivity and specificity values up to 94%. All the experiments were performed separately for RGB and HSV color spaces. Experimental results show the combination of the mean planes in red and green planes to be the best feature-subset in RGB (Red-Green-Blue) color space and the combination of the mean values of all three planes of the color space to be the best feature-subset in HSV (Hue-Saturation-Value).

On révise les prérequis de géométrie différentielle nécessaires à une première approche de la théorie de la quantification géométrique, c'est-à-dire des notions de base en géométrie symplectique, des notions de groupes et d'algèbres de Lie, d'action d'un groupe de Lie, de G-fibré principal, de connexion, de fibré associé et de structure presque-complexe. Ceci mène à une étude plus approfondie des fibrés en droites hermitiens, dont une condition d'existence de fibré préquantique sur une variété symplectique. Avec ces outils en main, nous commençons ensuite l'étude de la quantification géométrique, étape par étape. Nous introduisons la théorie de la préquantification, i.e. la construction des opérateurs associés à des observables classiques et la construction d'un espace de Hilbert. Des problèmes majeurs font surface lors de l'application concrète de la préquantification : les opérateurs ne sont pas ceux attendus par la première quantification et l'espace de Hilbert formé est trop gros. Une première correction, la polarisation, élimine quelques problèmes, mais limite grandement l'ensemble des observables classiques que l'on peut quantifier. Ce mémoire n'est pas un survol complet de la quantification géométrique, et cela n'est pas son but. Il ne couvre ni la correction métaplectique, ni le noyau BKS. Il est un à-côté de lecture pour ceux qui s'introduisent à la quantification géométrique. D'une part, il introduit des concepts de géométrie différentielle pris pour acquis dans (Woodhouse [21]) et (Sniatycki [18]), i.e. G-fibrés principaux et fibrés associés. Enfin, il rajoute des détails à quelques preuves rapides données dans ces deux dernières références. / We review some differential geometric prerequisite needed for an initial approach of the geometric quantization theory, i.e. basic notions in symplectic geometry, Lie group, Lie group action, principal G-bundle, connection, associated bundle, almost-complex structure. This leads to an in-depth study of Hermitian line bundles that leads to an existence condition for a prequantum line bundle over a symplectic manifold. With these tools, we start a study of geometric quantization, step by step. We introduce the prequantization theory, which is the construction of operators associated to classical observables and construction of a Hilbert space. Some major problems arise when applying prequantization in concrete examples : the obtained operators are not exactly those expected by first quantization and the constructed Hilbert space is too big. A first correction, polarization, corrects some problems, but greatly limits the set of classical observables that we can quantize. This dissertation is not a complete survey of geometric quantization, which is not its goal. It's not covering metaplectic correction, neither BKS kernel. It's a side lecture for those introducing themselves to geometric quantization. First, it's introducing differential geometric concepts taken for granted in (Woodhouse [21]) and (Sniatycki [18]), i.e. principal G-bundles and associated bundles. Secondly, it adds details to some brisk proofs given in these two last references.

We consider two different problems in quantization theory. During the first part we discuss the so called Bennett's White Noise Hypothesis, introduced to study quantization errors of different schemes. Under this hypothesis, one assumes that the reconstruction errors of different channels can be considered as uniform, independent and identically distributed random variables. We prove that in the case of uniform quantization errors for frame expansions, this hypothesis is in fact false. Nevertheless, we also prove that in the case of fine quantization, the errors of different channels are asymptotically uncorrelated, validating, at least partially, results on the computation of the mean square error of reconstructions that were obtained through the assumption of Bennett's hypothesis. On the second part, we will introduced a new scalar quantization scheme, called a Beta Alpha Encoder. We analyze its robustness with respect to the quantizer imperfections. This scheme also induces a challenging dynamical system. We give partial results dealing with the ergodicity of this system.

Thesis (Ph. D.)--University of Hawaii at Manoa, 1994. / Includes bibliographical references (leaves 159-166). / Microfiche. / xiv, 166 leaves, bound ill. 29 cm

In this dissertation, we study properties of quantum interaction phenomena in two-dimensional (2D) and one-dimensional (1D) electronic systems in p-GaAs micro- and nano-scale devices. We present low-temperature magneto-transport data from three forms of low-dimensional systems 1) 2D hole systems: in order to study interaction contributions to the metallic behavior of 2D systems 2) Bilayer hole systems: in order to study the many body, bilayer quantum Hall state at nu = 1 3) 1D hole systems: for the study of the anomalous conductance plateau G = 0.7 ???? 2e2/h The work is divided into five experimental studies aimed at either directly exploring the properties of the above three interaction phenomena or the development of novel device structures that exploit the strong particle-particle interactions found in p-GaAs for the study of many body phenomena. Firstly, we demonstrate a novel semiconductor-insulator-semiconductor field effect transistor (SISFET), designed specifically to induced 2D hole systems at a ????normal???? AlGaAs-on-GaAs heterojunction. The novel SISFETs feature in our studies of the metallic behavior in 2D systems in which we examine temperature corrections to ????xx(T) and ????xy(T) in short- and long-range disorder potentials. Next, we shift focus to bilayer hole systems and the many body quantum Hall states that form a nu = 1 in the presence of strong interlayer interactions. We explore the evolution of this quantum Hall state as the relative densities in the layers is imbalanced while the total density is kept constant. Finally, we demonstrate a novel p-type quantum point contact device that produce the most stable and robust current quantization in a p-type 1D systems to date, allowing us to observed for the first time the 0.7 structure in a p-type device.

Thesis (Ph. D.)--University of Washington, 1998. / Vita. Includes bibliographical references (leaf [106]).