This thesis focuses on implementation of an algorithm for retinal vessel segmentation in high resolution retinal images.A neural network with two hidden layers was used as the method. A total of 7 features were obtained from matched filtering based on vessel thickness, texture analysis and individual pixels brightness. Within the thesis, the whole database was manually annotated for the implementation of the algorithm and the results. The achieved mean sensitivity reached 80%, specificity 70% and Dice coefficient is 59%.

In frequency-domain photothermal radiometry (FD-PTR) a low-power intensity-modulated optical excitation generates thermal-wave field inside the sample and the subsequent infrared radiation from the sample is analyzed to detect material’s inhomogeneities. The non-contact nature of FD-PTR makes it very suitable for non-destructive evaluation of broad range of materials. Moreover, the methodology is based on intrinsic contrast of light absorption which can be used as a diagnostic tool for inspection of malignancy in biological tissues. Nevertheless, the bottom line is that the physics of heat diffusion allows for a highly damped and dispersive propagation of thermal-waves. As a result, the current FD-PTR modalities suffer from limited inspection depth and poor axial/depth resolution. The main objective of this thesis is to show that using alternative types of modulation schemes (such as linear frequency modulation and binary phase coding) and radar matched filter signal processing, one can obtain localized responses from inherently diffuse thermal wave fields. In this thesis, the photothermal responses of turbid, transparent, and opaque media to linear frequency modulated and binary phase coded excitations are analytically derived. Theoretical simulations suggest that matched-filtering in diffusion-wave field acts as constructive interferometry, localizing the energy of the long-duty excitation under a narrow peak and allowing one to construct depth resolved images. The developed technique is the diffusion equivalent of optical coherence tomography and is named thermal coherence tomography. It was found that the narrow-band binary phase coded matched filtering yields optimal depth resolution, while the broad-band linear frequency modulation can be used to quantify material properties through the multi-parameter fitting of the experimental data to the developed theory. Thermophotonic detection of early dental caries is discussed in detail as a potential diagnostic application of the proposed methodologies. The performance of the diagnostic system is verified through a controlled demineralization protocol as well as in teeth with natural caries.

In frequency-domain photothermal radiometry (FD-PTR) a low-power intensity-modulated optical excitation generates thermal-wave field inside the sample and the subsequent infrared radiation from the sample is analyzed to detect material’s inhomogeneities. The non-contact nature of FD-PTR makes it very suitable for non-destructive evaluation of broad range of materials. Moreover, the methodology is based on intrinsic contrast of light absorption which can be used as a diagnostic tool for inspection of malignancy in biological tissues. Nevertheless, the bottom line is that the physics of heat diffusion allows for a highly damped and dispersive propagation of thermal-waves. As a result, the current FD-PTR modalities suffer from limited inspection depth and poor axial/depth resolution. The main objective of this thesis is to show that using alternative types of modulation schemes (such as linear frequency modulation and binary phase coding) and radar matched filter signal processing, one can obtain localized responses from inherently diffuse thermal wave fields. In this thesis, the photothermal responses of turbid, transparent, and opaque media to linear frequency modulated and binary phase coded excitations are analytically derived. Theoretical simulations suggest that matched-filtering in diffusion-wave field acts as constructive interferometry, localizing the energy of the long-duty excitation under a narrow peak and allowing one to construct depth resolved images. The developed technique is the diffusion equivalent of optical coherence tomography and is named thermal coherence tomography. It was found that the narrow-band binary phase coded matched filtering yields optimal depth resolution, while the broad-band linear frequency modulation can be used to quantify material properties through the multi-parameter fitting of the experimental data to the developed theory. Thermophotonic detection of early dental caries is discussed in detail as a potential diagnostic application of the proposed methodologies. The performance of the diagnostic system is verified through a controlled demineralization protocol as well as in teeth with natural caries.

In this thesis, symbol timing recovery in MSK signals is investigated making use of matched filtering. A decision-directed symbol synchronizer cascaded with an MLSE receiver is proposed for fine timing. Correlation (matched filter) method is used to recover the timing epoch from the tentative decisions obtained from the Viterbi algorithm. The fractional delays are acquired using interpolation and an iterative maximum search process. In order to investigate the tracking performance of the proposed symbol synchronizer, a study is carried out on three possible optimum timing phase criteria: (i) Mazo criterion, (ii) the minimum squared ISI criterion (msISI), and (iii) the minimum BER criterion. Moreover, a discussion is given about the timing sensitivity of the MLSE receiver. The performance of the symbol synchronizer is assessed by computer simulations. It is observed that the proposed synchronizer tracks the variations of the channels almost the same as the msISI criterion. The proposed method eliminates the cycle slips very succesfully and is robust to frequency-selective multipath fading channel conditions even in moderate signal-to-noise ratios.

The purpose of this master thesis was to investigate how effiecient the extractionof radiofrequency quality metrics from digital video broadcast (DVB) streamscan become using software defined radio. Software defined radio (SDR) is a fairlynew technology that offers you the possibility of very flexible receivers and transmitters where it is possible to upgrade the modulation and demodulation overtime. Agama is interested in SDR for use in the Agama Analyzer, a widely deployedmonitoring probe running on top of standard services. Using SDR, Agama coulduse that in all deployments, such as DVB by cable/terrestrial/satellite (DVBC/T/S), which would simplify logistics. This thesis is an implementation of a SDR to be able to receive DVB-C. TheSDR must perform a number of adaptive algorithms in order to prevent the received symbols from being significantly different from the transmitted ones. Themain parts of the SDR include timing recovery, carrier recovery and equalization.Timing recovery performs synchronization between the transmitted and receivedsymbols and the carrier recovery performs synchronization between the carrierwave of the transmitter and the local oscillator in the receiver. The thesis discusses various methods to perform the different types of synchronizations andequalizations in order to find the most suitable methods.

This thesis further develops a method from ongoing thesis projects with the goal of generating images using synthetic aperture radar (SAR) simulations coded in MATLAB. The project is supervised by Dr. John Saghri and sponsored by Raytheon Space and Airborne Systems. SAR is a type of imaging radar in which the relative movement of the antenna with respect to the target is utilized. Through the simultaneous processing of the radar reflections over the movement of the antenna via the range Doppler algorithm (RDA), the superior resolution of a theoretical wider antenna, termed synthetic aperture, is obtained. The long term goal of this ongoing project is to develop a simulation in which realistic SAR images can be generated and used for SAR Automatic Target Recognition (ATR). Current and past Master’s theses on ATR were restricted to a small data set of Man-portable Surveillance and Target Acquisition Radar (MSTAR) images as most SAR images for military ATR are not released for public use. Also, with an in-house SAR image generation scheme the parameters of noise, target orientation, the elevation angle or look angle to the antenna from the target and other parameters can be directly controlled and modified to best serve ATR purposes or other applications such as three-dimensional SAR holography. At the start of the project in September 2007, the SAR simulation from previous Master’s theses was capable of simulating and imaging point targets in a two dimensional plane with limited mobility. The focus on improvements to this simulation through the course of this project was to improve the SAR simulation for applications to more complex two-dimensional targets and simple three-dimensional targets, such as a cube. The input to the simulation uses a selected two-dimensional, grayscale target image and generates from the input a two-dimensional target profile of reflectivity over the azimuth and range based on the intensity of the pixels in the target image. For three-dimensional simulations, multiple two-dimensional azimuth/range profiles are imported at different altitudes. The output from both the two-dimensional and three-dimensional simulations is the SAR simulated and RDA processed image of the input target profile. Future work on this ongoing project will include an algorithm to calculate line of sight limitations of point targets and processing optimization of the radar information generation implemented in the code so that more complex and realistic targets can be simulated and imaged using SAR for applications in ATR and 3D SAR holography.

La tesi tracta, des del punt de vista de l’anàlisi de dades, la possibilitat de detecció directa d’ones gravitatòries emeses per sistemes binaris d’objectes compactes de massa similar: forats negres, estels de neutrons, nanes blanques. En els capítols introductoris, a) es dóna una descripció detallada i exhaustiva de com passar dels patrons d’ona teòrics a la senyal detectada; b) s’introdueixen les eines més emprades en l’anàlisi de dades d’ones gravitatòries, amb especial menció a la discussió sobre les amplituds efectiva i característica. A més, els resultats originals de la tesi segueixen tres línies de recerca diferents: 1) S’ha predit la precisió amb la que el futur detector interferomètric espacial LISA, estimarà els paràmetres (posició, masses, velocitat de rotació, paràmetres cosmològics…) de les observacions de xocs entre dos forats negres supermassius en la fase “inspiral”. 2) S’ha desenvolupat un algorisme propi de cerca de senyals gravitatòries procedents de sistemes binaris estel•lars, basat en teories de probabilitat Bayesiana i MCMC. Aquest algorisme distingeix alhora milers de senyals superposades en una única sèrie temporal de dades, extraient paràmetres individuals de cadascuna d’elles. 3) S’ha definit de manera matemàtica rigorosa com determinar el rang de validesa (per a extracció de paràmetres i detecció) de models aproximats de patrons d’ones gravitatòries, aplicant-ho a un cas concret de models semi-analítics / La tesis trata, desde el punto de vista del análisis de datos, la posibilidad de detección directa de ondas gravitacionales emitidas por sistemas binarios de objetos compactos de masa similar: agujeros negros, estrellas de neutrones, enanas blancas. En los capítulos introductorios, a) se desarrolla una descripción detallada y exhaustiva de como pasar de los patrones de onda teóricos a la señal detectada; b) se introducen las herramientas más utilizadas en el análisis de datos de ondas gravitacionales, con especial mención a la discusión sobre las amplitudes efectiva y característica. Además, los resultados originales de la tesis siguen tres líneas de investigación diferentes: 1) Se ha predicho la precisión con la que el futuro detector interferométrico espacial LISA, estimará los parámetros (posición, masas, velocidad de rotación, parámetros cosmológicos…) de las observaciones de choques entre dos agujeros negros supermasivos en la fase “inspiral”. 2) Se ha desarrollado un algoritmo propio de búsqueda de señales gravitacionales procedentes de sistemas binarios estelares, basado en teorías de probabilidad Bayesiana y MCMC. Este algoritmo distingue a la vez miles de señales superpuestas en una única serie temporal de datos, extrayendo parámetros individuales de cada una de ellas. 3) Se ha definido de manera matemática rigurosa como determinar el rango de validez (para extracción de parámetros y detección) de modelos aproximados de patrones de ondas gravitacionales, aplicándolo a un caso concreto de modelos semi-analíticos. / In this PhD thesis one studies, from the data analysis perspective, the possibility of direct detection of gravitational waves emitted by similar mass compact binary objects: black holes, neutron stars, white dwarfs. In the introductory chapters, a) a detailed and exhaustive description about how to derive the detected strain from the theoretical emitted waveform predictions is given; b) the most used gravitational wave data analysis results are derived, being worth pointing out the discussion about effective and characteristic amplitudes. Moreover, three different research lines have been followed in the thesis: 1) It has been predicted the parameter estimation (position, masses, spin, cosmological parameters…) of supermassive black hole binary inspiral signals, observed with the future interferometric space detector, LISA. 2) A new algorithm, based on Bayesian probability and MCMC techniques, has been developed in order to search for gravitational wave signals from stellar-mass binary systems. The algorithm is able to distinguish thousands of overlapping signals from a single observed time series, allowing for individual parameter extraction. 3) It has been, mathematically and rigorously, defined how to compute the validity range (for parameter estimation and detection purposes) of approximated gravitational waveform models, applying it to the particular case of closed-form models

In this thesis, symbol timing recovery based on matched filtering in Gaussian Minimum Shift Keying (GMSK) with bandwidth-bit period product (BT) of 0.3 is investigated. GMSK is the standard modulation type for GSM. Although GMSK modulation is non-linear, it is approximated to Offset Quadrature Amplitude Modulation (OQAM), which is a linear modulation, so that Maximum Likelihood Sequence Estimation (MLSE) method is possible in the receiver part. In this study Typical Urban (TU) channel model developed in COST 207 is used. Two methods are developed on the construction of the matched filter. In order to obtain timing recovery for GMSK signals, these methods are investigated. The fractional time delays are acquired by using interpolation and an iterative maximum search process. The performance of the proposed symbol timing recovery (STR) scheme is assessed by using computer simulations. It is observed that the STR tracks the variations of the frequency selective multipath fading channels almost the same as the Mazo criterion.

Le sujet porte sur l'étude de systèmes SAS (Synthetic Aperture Sonar) multistatiques. Ces systèmes permettent d'obtenir des images de cibles mieux résolues qu'avec un sonar classique à partir d'ondes acoustiques. Le SAS est largement exploité en configuration monostatique mais il existe très peu d'études à ce jour en SAS multistatique. Le travail consiste donc à évaluer les performances en configuration bistatique et multistatique et à les comparer à celles connues en configuration monostatique. Une méthode de calcul utilisée en radar a donc été mise en oeuvre en sonar de façon à expliciter la résolution en configuration bistatique, ce qui est un résultat original de ce travail. L'algorithme classiquement utilisé pour reconstruire des images repose sur l'hypothèse que la cible est une somme de points brillants. Cette hypothèse n'est pas bien adaptée en acoustique sous-marine. Un nouvel algorithme a donc été développé dans le but de se rapprocher des phénomènes de diffraction présents à l'interface entre l'eau et la cible. Le modèle de champ diffracté est obtenu par la combinaison d'équations intégrales de frontière avec l'approximation de Kirchhoff. Une méthode de reconstruction d'images par transformée de Fourier 2D de ce modèle a été implémentée et testée sur des données simulées, puis sur des données obtenues lors d'essais en cuve. Le nouvel algorithme montre une meilleure précision de la reconstruction et la capacité de pouvoir extraire de l'information quantitative de la cible. L'intérêt des configurations multistatiques pour la reconnaissance de cibles a également été démontré dans ces travaux de thèse. / This study deals with multistatic SAS (Synthetic Aperture Sonar) systems. SAS are high resolution imaging systems compared to classical sonar ones. The SAS technique is largly exploited in the monostatic configuration but few studies already exist in multistatic SAS. Thus, the work consists in evaluating resolution and detection performances in bistatic and multistatic configurations. Then, the objective is to compare these performances to monostatic ones. A radar method has been adapted to sonar to compute bistatic performances and this is an original result of this work.The classical algorithm to reconstruct images from acoustical waves lies on the hypothesis that the target is a sum of point scatterers. This hypothesis is not really well adapted to underwater acoustics that is why a new algogorithm has been developped in this study. The new algorithm would be better adapted to scattering diffraction phenomena at the interface between water and target than the classical one. The scattered field model of the target is obtained by combinating boundary integral equations and the Kirchhoff Approximation. An imaging reconstruction method by 2D Fourier Transform of this model has been implemented and tested on numerical and experimental datas. The new algorithm allow a better reconstruction accurency and is able to give quantitative information on targets. The interest of multistatic configurations for target identification has also been demonstrated in this PhD work.

Segmentation of vessel structure is an important phase in analysis of retinal images. The resulting vessel system description may be important for diagnostic of many eye and cardiovascular diseases. A method for automatic segmentation of the vessel structure in colour retinal images is presented in the thesis. The method utilises 2D matched filtering to detect presence of short linear vessel sections of a particular thickness and orientation. The approach correlates the local image areas with a 2D masks based on a typical brightness profile perpendicular to vessels of a particular width. Three different approximated profiles are used and corresponding matched filters are designed for: thin, medium and thick vessels. The evaluation of typical vessel profiles and filter design are described in chapter 3 and chapter 4. The parametric images obtained by convolution of the image with the masks are then thresholded in order to obtain binary representation of vessel structure. The three binary representations are consequently combined to provide the best available rough vessel map, which is finalised by complementing the obviously missing vessel sections and cleaning the disconnected fractional artefacts. The thresholding algorithm and final steps of processing are mentioned in chapter 5 and chapter 6. The method has been implemented by computer and the program for automatic vessel segmentation has been developed using database of real retinal images. The efficiency of the method has been finally evaluated on images from the standard database DRIVE.