Over the past several decades, push-broom imaging spectrometers have become a common Earth observation tool. Instruments of this type must be calibrated to convert the raw sensor data into units of spectral radiance. Calibration is in this case a two-step process: First, a sensor model is obtained by performing calibration measurements, which is then used to convert raw signals to spectral radiance data. Further processing steps can be performed to correct for optical image distortions. In this work, we show the complete calibration process for push-broom imaging spectrometers, including uncertainty propagation. Although the focus is specifically on calibrating a HySpex VNIR-1600 airborne-imaging spectrometer, all methods can be adapted for other instruments. We discuss the theory of push-broom imaging spectrometers by introducing a generic sensor model, which includes the main parameters and effects of such instruments. Calibrating detector-related effects, such as dark signal, the noise as a function of the signal, and temperature effects is shown. Correcting temperature effects significantly reduces measurement errors. To determine the signal non-linearity, we built a setup based on the light-addition method and improved this method to allow smaller signal level distances of the sampling points of the non-linearity curve. In addition, we investigate the non-linearity of the integration time. The signal (<=15%) and the integration time (<=0.5%) non-linearities can be corrected with negligible errors. After correcting both non-linearity effects, a smearing effect is revealed, which is investigated in detail. We use a collimator and monochromator setup for calibrating the geometric and spectral parameters, respectively. To accurately model the angular and spectral response functions, we propose using cubic splines, which leads to significant improvements compared to previously used Gaussian functions. We present a new method that allows interpolation of the cubic spline based response functions for pixels not measured. The results show that the spectral and geometric properties are non-uniform and change rapidly within a few pixels. The absolute radiometric calibration is performed with a lamp-plaque setup and an integrating sphere is used for flat-fielding. To mitigate the influence of sphere non-uniformities, we rotate the instrument along the across-track angle to measure the same spot of the sphere with each pixel. We investigate potential systematic errors and use Monte Carlo simulations to determine the uncertainties of the radiometric calibration. In addition, we measure the polarization sensitivity with a wire-grid polarizer. Finally, we propose a novel image transformation method that allows manipulation of geometric and spectral properties of each pixel individually. Image distortions can be corrected by changing a pixel's center angles, center wavelength, and response function shape. This is done by using a transformation matrix that maps each pixel of a target sensor B to the pixels of a source sensor A. This matrix is derived from two cross-correlation matrices: Sensor A and itself, and sensor B and sensor A. We provide the mathematical background and discuss the propagation of uncertainty. A case study shows that the method can significantly improve data quality.

Railroad tracks require consistent and periodic monitoring to ensure safety and reliability. Unmanned Aerial Vehicles (UAVs) have great potential because they are not constrained to the track, allowing trains to continue running while the UAV is inspecting. Also, they can be quickly deployed without human intervention. For these reasons, the first steps towards creating a track-monitoring UAV system have been completed. This thesis focuses on the design of algorithms to be deployed on a UAV for the purpose of monitoring the health of railroad tracks. Before designing the algorithms, the first steps were to design a rough physical structure of the final product. A small multirotor or fixed-wing UAV will be used with a gimbaled camera mounted on the belly. The camera will take images of the tracks while the onboard computer processes the images. The computer will locate the tracks in the image as well as perform defect detection on those tracks. Algorithms were implemented once a rough physical structure of the product was completed. These algorithms detect and follow rails through a video feed and detect defects in the rails. The rail following algorithm is based on a custom-designed masking technique that locates rails in images. A defect detection algorithm was also created. This algorithm detect defects by analyzing gradient data on the rail surface. / Master of Science

This master's thesis deals with acceleration of pixel interpolation methods using the GPU and NVIDIA (R) CUDA TM architecture. Graphic output is represented by a demonstrational application for geometrical image transforms using chosen interpolation method. Time critical parts of the code are moved on the GPU and executed in parallel. There are used highly optimized routines from the OpenCV library, made by the Intel company for an image and video processing.

Using presentations in an m-Learning environment enables delivery of rich content to a mobile phone learner. This study investigated how to prepare and stream presentations from a desktop computer to a mobile phone in near-realtime. It also addressed communication between users using interactive text chat in the same environment. Our analysis of text/podcast-based m-Learning applications revealed limited interactivity and lack of diversity in content streamed. To address this, we developed a mobile-based application that uses a task-timer model to synchronize with a server every n units of time to enable near-realtime streaming of presentation slides between mobile and desktop users. The application included text-based instant messenger. Laboratory experiments investigated the use OpenOffice and PowerPoint presentations and techniques used to convert these presentations into mobile phone compatible formats. Experiments were carried out with smart mobile phones running on a third generation cellular network. We employed transaction-logging techniques in addition to automated image analysis techniques to observe and record data. Analysis of the results revealed using presentations enabled more rich content than text-based models such as short message service-based frameworks and podcasts. Although m-Learning is not yet widely adopted, applications such as the one developed in the study offer high hopes for m-Learning because of the use of rich content and interactivity between users.

Using presentations in an m-Learning environment enables delivery of rich content to a mobile phone learner. This study investigated how to prepare and stream presentations from a desktop computer to a mobile phone in near-realtime. It also addressed communication between users using interactive text chat in the same environment. Our analysis of text/podcast-based m-Learning applications revealed limited interactivity and lack of diversity in content streamed. To address this, we developed a mobile-based application that uses a task-timer model to synchronize with a server every n units of time to enable near-realtime streaming of presentation slides between mobile and desktop users. The application included text-based instant messenger. Laboratory experiments investigated the use OpenOffice and PowerPoint presentations and techniques used to convert these presentations into mobile phone compatible formats. Experiments were carried out with smart mobile phones running on a third generation cellular network. We employed transaction-logging techniques in addition to automated image analysis techniques to observe and record data. Analysis of the results revealed using presentations enabled more rich content than text-based models such as short message service-based frameworks and podcasts. Although m-Learning is not yet widely adopted, applications such as the one developed in the study offer high hopes for m-Learning because of the use of rich content and interactivity between users.

Masters of Science / Using presentations in an m-Learning environment enables delivery of rich content to a mobile phone learner. This study investigated how to prepare and stream presentations from a desktop computer to a mobile phone in near-realtime. It also addressed communication between users using interactive text chat in the same environment. Our analysis of text/podcast-based m-Learning applications revealed limited interactivity and lack of diversity in content streamed. To address this, we developed a mobile-based application that uses a task-timer model to synchronize with a server every n units of time to enable near-real time streaming of presentation slides between mobile and desktop users. The application included text-based instant messenger. Laboratory experiments investigated the use Open Office and PowerPoint presentations and techniques used to convert these presentations into mobile phone compatible formats. Experiments were carried out with smart mobile phones running on a third generation cellular network. We employed transaction-logging techniques in addition to automated image analysis techniques to observe and record data. Analysis of the results revealed using presentations enabled more rich content than text-based models such as short message service-based frameworks and podcasts. Although m-Learning is not yet widely adopted, applications such as the one developed in the study offer high hopes for m-Learning because of the use of rich content and interactivity between users. / South Africa

Nous présentons dans cette étude nos travaux concernant le recalage 3D/2D d'images de dissection aortique. Son but est de de proposer une visualisation de données médicales, qui pourra servir dans le contexte de l'assistance peropératoire durant les procédures endovasculaires.Pour effectuer cette tâche, nous avons proposé un modèle paramétrique de l'aorte, appelé enveloppe tubulaire. Il sert à exprimer la forme globale et les déformations de l'aorte, à l'aide d'un nombre minimal de paramètres.L'enveloppe tubulaire est utilisée par les algorithmes de recalage proposés dans cette étude.Notre méthode originale consiste à proposer un recalage par calcul direct de la transformation entre image 2D, i.e. sans procéssus d'optimisation, et est appelée recalage par ITD .Les descripteurs, que nous avons définis pour le cas des images d'aorte, permettent de trouver rapidement un alignement grossier des données. Nous proposons également l'extension de notre approche pour la mise en correspondance des images 3Det 2D.La chaîne complète du recalage 3D/2D, que nous présentons dans ce document, est composée de la technique ITD et de méthodes précises iconiques et hybrides. L'intégration de notre algorithme basé sur les descripteurs en tant qu'étape d'initialisation réduit le temps de calcul nécessaire et augmente l'efficacité du recalage, par rapport aux approches classiques.Nous avons testé nos méthodes avec des images médicales, issues de patients trîtés par procédures endovasculaires. Les résultats ont été vérifiés par les spécialistes cliniques et ont été jugés satisfaisants; notre chaine de recalage pourrait ainsi être exploitée dans les salles d'interventions à l'avenir. / In this study, we present our works related to 3D/2D image registrationfor aorti dissition. Its aim is to propose a visualization of medial datawhih an be used by physians during endovas ular proedures.For this purpose, we have proposed a parametrimodel of aorta, alleda Tubular Envelope. It is used to express the global shape and deformationsof the aorta, by a minimal number of parameters. The tubular envelope isused in our image registration algorithms.The registration by ITD (Image Transformation Descriptors) is our ori-ginal method of image alignment : itomputes the rigid 2D transformation between data sets diretly, without any optimization process.We provide thedefinition of this method, as well as the proposition of several descriptors' formulae, in the base of images of aorta. The technique allows us to quickly and a poarse alignment between data. We also propose the extension of theoriginal approach for the registration of 3D and 2D images.The complete chain of 3D/2D image registration techniques, proposedin this document, consists of the ITD stage, followed by an intensity basedhybrid method. The use of our 3D/2D algorithm, based on the image trans-formation descriptors as an initialization phase, reduces the computing timeand improves the efficiency of the presented approach.We have tested our registration methods for the medical images of several patients after endovasular treatment. Results have been approved by our clinical specialists and our approach.We have tested our registration methods for the medical images of several patients after endovascular treatment. Results have been approved by our clinical specialists and our approach may appear in the intervention rooms in the futur.

Fúze obrazu je dnes jednou z nejběžnějších avšak stále velmi diskutovanou oblastí v lékařském zobrazování a hraje důležitou roli ve všech oblastech lékařské péče jako je diagnóza, léčba a chirurgie. V této dizertační práci jsou představeny tři projekty, které jsou velmi úzce spojeny s oblastí fúze medicínských dat. První projekt pojednává o 3D CT subtrakční angiografii dolních končetin. V práci je využito kombinace kontrastních a nekontrastních dat pro získání kompletního cévního stromu. Druhý projekt se zabývá fúzí DTI a T1 váhovaných MRI dat mozku. Cílem tohoto projektu je zkombinovat stukturální a funkční informace, které umožňují zlepšit znalosti konektivity v mozkové tkáni. Třetí projekt se zabývá metastázemi v CT časových datech páteře. Tento projekt je zaměřen na studium vývoje metastáz uvnitř obratlů ve fúzované časové řadě snímků. Tato dizertační práce představuje novou metodologii pro klasifikaci těchto metastáz. Všechny projekty zmíněné v této dizertační práci byly řešeny v rámci pracovní skupiny zabývající se analýzou lékařských dat, kterou vedl pan Prof. Jiří Jan. Tato dizertační práce obsahuje registrační část prvního a klasifikační část třetího projektu. Druhý projekt je představen kompletně. Další část prvního a třetího projektu, obsahující specifické předzpracování dat, jsou obsaženy v disertační práci mého kolegy Ing. Romana Petera.

Etude bibliographique sur le recalage d'images de visage et sur le recalage d'images et travail en collaboration avec Son VuS, pour définir la précision nécessaire du recalage en fonction des exigences des méthodes de reconnaissance de visages. / Face alignment is an important step in a typical automatic face recognition system.This thesis addresses the alignment of faces for face recognition applicationin video surveillance context. The main challenging factors of this research includethe low quality of images (e.g., low resolution, motion blur, and noise), uncontrolledillumination conditions, pose variations, expression changes, and occlusions. In orderto deal with these problems, we propose several face alignment methods using differentstrategies. The _rst part of our work is a three-stage method for facial pointlocalization which can be used for correcting mis-alignment errors. While existingalgorithms mostly rely on a priori knowledge of facial structure and on a trainingphase, our approach works in an online mode without requirements of pre-de_nedconstraints on feature distributions. The proposed method works well on images underexpression and lighting variations. The key contributions of this thesis are aboutjoint image alignment algorithms where a set of images is simultaneously alignedwithout a biased template selection. We respectively propose two unsupervised jointalignment algorithms : \Lucas-Kanade entropy congealing" (LKC) and \gradient correlationcongealing" (GCC). In LKC, an image ensemble is aligned by minimizing asum-of-entropy function de_ned over all images. GCC uses gradient correlation coef-_cient as similarity measure. The proposed algorithms perform well on images underdi_erent conditions. To further improve the robustness to mis-alignments and thecomputational speed, we apply a multi-resolution framework to joint face alignmentalgorithms. Moreover, our work is not limited in the face alignment stage. Since facealignment and face acquisition are interrelated, we develop an adaptive appearanceface tracking method with alignment feedbacks. This closed-loop framework showsits robustness to large variations in target's state, and it signi_cantly decreases themis-alignment errors in tracked faces.

One of the main problems in neurosurgery is knowledge about human brain. It's very important to see the whole brain with its critical neurostructures in virtual reality. This document is about three dimensional human brain visualization strategies. Review most recently used three dimensional objects building strategies from two dimensional medical MRI images. This task was split into 4 significant problems: image segmentation, point-sets correspondence, image registration and its frequently used transformation functions with image matching measurements. All these problems were addressed reader to show most recently used algorithms with advantages and disadvantages. Atlas types, patterns and maps survey was introduced with widely popular brain model coordinate systems. In order to find a better correspondence between two point sets it was modeled a new robust and accurate Overhauser spline points location optimization algorithm. Instead of deletion outlier points from overloaded point set, this algorithm generates more points in other set at optimized points locations. Determination of an accurate point location and choosing the correct transformation function are the key steps in registration process. Whereas registration is vital task in precise human brain visualization for neurosurgeries at preoperative and intraoperative process.