In this dissertation, a fully scalable video coding system is proposed. This system achieves full temporal, resolution, and fidelity scalability by combining mesh-based motion-compensated temporal filtering, multihypothesis motion compensation, and an embedded 3D wavelet-coefficient coder. The first major contribution of this work is the introduction of the redundant-wavelet multihypothesis paradigm into motion-compensated temporal filtering, which is achieved by deploying temporal filtering in the domain of a spatially redundant wavelet transform. A regular triangle mesh is used to track motion between frames, and an affine transform between mesh triangles implements motion compensation within a lifting-based temporal transform. Experimental results reveal that the incorporation of redundant-wavelet multihypothesis into mesh-based motion-compensated temporal filtering significantly improves the rate-distortion performance of the scalable coder. The second major contribution is the introduction of a sliding-window implementation of motion-compensated temporal filtering such that video sequences of arbitrarily length may be temporally filtered using a finite-length frame buffer without suffering from severe degradation at buffer boundaries. Finally, as a third major contribution, a novel 3D coder is designed for the coding of the 3D volume of coefficients resulting from the redundant-wavelet based temporal filtering. This coder employs an explicit estimate of the probability of coefficient significance to drive a nonadaptive arithmetic coder, resulting in a simple software implementation. Additionally, the coder offers the possibility of a high degree of vectorization particularly well suited to the data-parallel capabilities of modern general-purpose processors or customized hardware. Results show that the proposed coder yields nearly the same rate-distortion performance as a more complicated coefficient coder considered to be state of the art.

L'Imagerie par Résonance Magnétique de la cinématique de la contraction cardiaque est une technique d'imagerie relativement lente. En comparaison, les mouvements du patient, en particulier cardiaque et respiratoire, sont rapides et peuvent provoquer des artéfacts sur les images. La vitesse de contraction cardiaque apporte justement des informations cliniquement utiles. Premièrement, nous avons montré qu'il était possible d'effectuer cette mesure en IRM Cine à contraste de phase, et d'obtenir des valeurs similaires à celles obtenues de façon clinique en échographie cardiaque. La condition est d'obtenir une haute résolution temporelle, or, pour ce faire, la durée d'acquisition doit être plus longue qu'une apnée. La gestion du mouvement respiratoire en respiration libre a été réalisée de deux façons : avec moyennage puis avec correction de mouvement à l'aide de Cine-GRICS. Deuxièmement, pour atteindre une bonne reconstruction de la résolution temporelle en Cine, nous avons proposé une gestion temps réel de la variation du rythme cardiaque pendant l'acquisition IRM Cine, avec la construction d'un modèle cardiaque adapté au patient à l'aide de l'IRM à contraste de phase temps réel. Enfin, la gestion du mouvement cardio-respiratoire en IRM Cine est appliquée chez le petit animal à l'aide d'écho navigateurs IntraGate / Cine MRI of cardiac contraction is a relatively slow imaging technique. Comparatively, patient motion, especially cardiac beating and breathing, are fast and can lead to imaging artefacts. Cardiac contraction velocity provides clinically useful information. Firstly, we have shown that making this measurement was possible using phase contrast Cine MRI, and that getting similar values as those obtained in clinical routine using cardiac echography. The condition is to reach high temporal resolution, but to do so, the acquisition duration must be longer than a breathhold. Free-breathing motion management was done by two approaches: by averaging then by motion compensation using Cine-GRICS. Secondly, in order to achieve high temporal resolution Cine reconstruction, we proposed a way to deal with changing heart rate during Cine MRI acquisition, by the construction of a patient adapted cardiac model using realtime phase contrast MRI. Finally, cardio-respiratory motion management was adapted to small animal Cine MRI thanks to IntraGate echo navigators

Four dimensional imaging has become part of the standard of care for diagnosing and treating non-small cell lung cancer. In radiotherapy applications 4D fan-beam computed tomography (4D-CT) and 4D cone-beam computed tomography (4D-CBCT) are two advanced imaging modalities that afford clinical practitioners knowledge of the underlying kinematics and structural dynamics of diseased tissues and provide insight into the effects of regular organ motion and the nature of tissue deformation over time. While these imaging techniques can facilitate the use of more targeted radiotherapies, issues surrounding image quality and accuracy currently limit the utility of these images clinically. The purpose of this project is to develop methods that retrospectively compensate for anatomical motion in 4D-CBCT and correct motion artifacts present in 4D-CT to improve the image quality of reconstructed volume and assist in localizing respiration-influenced, diseased tissue and mobile structures of interest. In the first half of the project, a series of motion compensation (MoCo) workflow methods incorporating groupwise deformable image registration and projection-warped reconstruction were developed for use with 4D-CBCT imaging. In the latter half of the project, novel motion artifact observation and artifact- weighted groupwise registration-based image correction algorithms were designed and tested. Both deliverable components of this project were evaluated for their ability to enhance image quality when applied to clinical patient datasets and demonstrated qualitative and quantitative improvements over current state-of-the-art.

Context: Video stabilization techniques have gained popularity for their permit to obtain high quality video footage even in non-optimal conditions. There have been significant works done on video stabilization by developing different algorithms. Most of the stabilization software displays the missing image areas in stabilized video. In the last few years hand-held video cameras have continued to grow in popularity, allowing everyone to easily produce personal video footage. Furthermore, with online video sharing resources being used by a rapidly increasing number of users, a large proportion of such video footage is shared with wide audiences. Sadly such videos often suffer from poor quality as frame vibration in video makes human perception not comfortable. In this research an attempt has been made to propose a robust video stabilization algorithm that stabilizes the videos effectively. Objectives: The main objective of our thesis work is to perform effective motion estimation using SIFT features to calculate the inter frame motion, allowing to find Global Motion Vectors and optimal motion compensation is to be achieved using adaptive fuzzy filter by removing the unwanted shakiness and preserve the panning leading to stabilized video. Methods: In this study three types of research questions are used- Experimentation and Literature review. To accomplish the goal of this thesis, experimentation is carried out for performing video stabilization. Motion estimation is done using feature based motion estimation using SIFT and GMVs are calculated. The intentional motion is filtered using Adaptive fuzzy filter to preserve panning and Motion compensation is performed to wrap the video to its stabilized position. MOS implies the mean scores of the subjective tests performed according to the recommendations of ITU-R BT.500-13 and ITU-T P.910 to analyze the results of our stabilized videos. Results: As a part of results from our work, we have successfully stabilized the videos of different resolutions from experimentation. Performance of our algorithm is found better using MOS. Conclusions: Video Stabilization can be achieved successfully by using SIFT features with pre conditions defined for feature matching and attempts are made to improve the video stabilization process.

This thesis deals with description of lossy video compression. Theoretical part of the work describes the fundamentals of the video compression and standarts for lossy as well lossless video and still image compression. The practical part follows up with design of Java program for simulation of MPEG codec.

Le but de la radiothérapie est d’irradier les cellules cancéreuses tout en préservant au maximum les tissus sains environnants. Or, dans le cas du cancer du poumon, la respiration du patient engendre des mouvements de la tumeur pendant le traitement. Une solution possible est de repositionner continuellement le faisceau d’irradiation sur la cible tumorale en mouvement. L’e cacité et la sûreté de cette approche reposent sur la localisation précise en temps réel de la tumeur. Le suivi indirect consiste à inférer la position de la cible tumorale à partir de l’observation d’un signal substitut, visible en continu sans nécessiter de rayonnement ionisant. Un modèle de corrélation spatial doit donc être établi. Par ailleurs, pour compenser la latence du système, l’algorithme de suivi doit pouvoir également anticiper la position future de la cible. Parce que la respiration du patient varie dans le temps, les modèles de prédiction et de corrélation peuvent devenir imprécis. La prédiction de la position de la tumeur devrait alors idéalement être complétée par l’estimation des incertitudes associées aux prédictions. Dans la pratique clinique actuelle, ces incertitudes de positionnement en temps réel ne sont pas explicitement prédites. Cette thèse de doctorat s’intéresse au contrôle en temps réel de la précision du suivi indirect de tumeurs mobiles en radiothérapie. Dans un premier temps, une méthode bayésienne pour le suivi indirect en radiothérapie est développée. Cette approche, basée sur le filtre de Kalman, permet de prédire non seulement la position future de la tumeur à partir d’un signal substitut, mais aussi les incertitudes associées. Ce travail o re une première preuve de concept, et montre également le potentiel du foie comme substitut interne, qui apparait plus robuste et fiable que les marqueurs externes communément utilisés dans la pratique clinique. Dans un deuxième temps, une adaptation de la méthode est proposée afin d’améliorer sa robustesse face aux changements de respiration. Cette innovation permet de prédire des régions de confiance adaptatives, capables de détecter les erreurs de prédiction élevées, en se basant exclusivement sur l’observation du signal substitut. Les résultats révèlent qu’à sensibilité élevée (90%), une spécificité d’environ 50% est obtenue. Un processus de validation innovant basé sur ces régions de confiance adaptatives est ensuite évalué et comparé au processus conventionnel qui consiste en des mesures de la cible à intervalles de temps fixes et prédéterminés. Une version adaptative de la méthode bayésienne est donc développée afin d’intégrer des mesures occasionnelles de la position de la cible. Les résultats confirment que les incertitudes prédites par la méthode bayésienne permettent de détecter les erreurs de prédictions élevées, et démontrent que le processus de validation basé sur ces incertitudes a le potentiel d’être plus e cace que les validations régulières. Ces approches bayésiennes sont validées sur des séquences respiratoires de volontaires, acquises par imagerie par résonance magnétique (IRM) dynamique et interpolées à haute fréquence. Afin de compléter l’évaluation de la méthode bayésienne pour le suivi indirect, une validation expérimentale préliminaire est conduite sur des données cliniques de patients atteints de cancer du poumon. Les travaux de ce projet doctoral promettent une amélioration du contrôle en temps réel de la précision des prédictions lors des traitements de radiothérapie. Finalement, puisque l’imagerie ultrasonore pourrait être employée pour visualiser les substituts internes, une étude préliminaire sur l’évaluation automatique de la qualité des images ultrasonores est présentée. Ces résultats pourront être utilisés ultérieurement pour le suivi indirect en radiothérapie en vue d’optimiser les acquisitions ultrasonores pendant les traitements et faciliter l’extraction automatique du mouvement du substitut. / The goal of radiotherapy is to irradiate cancer cells while maintaining a low dose of radiation to the surrounding healthy tissue. In the case of lung cancer, the patient’s breathing causes the tumor to move during treatment. One possible solution is to continuously reposition the irradiation beam on the moving target. The e ectiveness and safety of this approach rely on accurate real-time localization of the tumor. Indirect strategies derive the target positions from a correlation model with a surrogate signal, which is continuously monitored without the need for radiation-based imaging. In addition, to compensate for system latency, the tracking algorithm must also be able to anticipate the future position of the target. Because the patient’s breathing varies over time, prediction and correlation models can become inaccurate. Ideally, the prediction of the tumor location would also include an estimation of the uncertainty associated with the prediction. However, in current clinical practice, these real-time positioning uncertainties are not explicitly predicted. This doctoral thesis focuses on real-time control of the accuracy of indirect tracking of mobile tumors in radiotherapy. First, a Bayesian method is developed. This approach, based on Kalman filter theory, allows predicting both future target motion in real-time from a surrogate signal and associated uncertainty. This work o ers a first proof of concept, and also shows the potential of the liver as an internal substitute as it appears more robust and reliable than the external markers commonly used in clinical practice. Second, an adaptation of the method is proposed to improve its robustness against changes in breathing. This innovation enables the prediction of adaptive confidence regions that can be used to detect significant prediction errors, based exclusively on the observation of the surrogate signal. The results show that at high sensitivity (90%), a specificity of about 50% is obtained. A new validation process based on these adaptive confidence regions is then evaluated and compared to the conventional validation process (i.e., target measurements at fixed and predetermined time intervals). An adaptive version of the Bayesian method is therefore developed to valuably incorporate occasional measurements of the target position. The results confirm that the uncertainties predicted by the Bayesian method can detect high prediction errors, and demonstrate that the validation process based on these uncertainties has the potential to be more e cient and e ective than regular validations. For these studies, the proposed Bayesian methods are validated on respiratory sequences of volunteers, acquired by dynamic MRI and interpolated at high frequency. In order to complete the evaluation of the Bayesian method for indirect tracking, experimental validation is conducted on clinical data of patients with lung cancer. The work of this doctoral project promises to improve the real-time control of the accuracy of predictions during radiotherapy treatments. Finally, since ultrasound imaging could be used to visualize internal surrogates, a preliminary study on automatic ultrasound image quality assessment is presented. These results can later be used for indirect tracking in radiotherapy to optimize ultrasound acquisitions during treatments and facilitate the automatic estimation of surrogate motion.

Super-resolution is a subfield of computer vision centered around upsampling low-resolution images to a corresponding high-resolution counterpart. This degree project investigates the suitability of a deep learning method for real-time video super-resolution. Following earlier work in the field, we use bicubic interpolation as a baseline for comparison. The deep learning method selected is specifically suited towards real-time super-resolution and consists of a motion compensation network and an upsampling network. The deep learning method and bicubic interpolation are compared by quantitatively evaluating the methods against each other in quality metrics and performance metrics. Suitable quality metrics are selected from earlier works to provide increased comparability of results, namely peak signal-to-noise ratio and structure similarity index. The performance metrics are: number of operations for a single upsampled frame, latency, throughput, and memory requirements. We apply the methods to a highly challenging publicly available dataset specifically engineered towards video super-resolution research. To further investigate the deep learning method, we propose a few modifications and study the effect on the metrics. Our findings show that the deep learning models outperform bicubic interpolation in the quality metrics, while bicubic interpolation outperformed the deep learning models in the performance metrics. We also find no significant quality metric improvement associated with having a motion compensation network for this dataset, suggesting that the dataset might be too complex for the motion compensation network. We conclude that the deep learning method exhibits real-time capabilities as the method has a throughput of around 500 frames per second for full HD super-resolution. Additionally, we show that by modifying the deep learning method, we achieve similar latency as bicubic interpolation without sacrificing throughput or quality. / Superupplösning är ett område inom datorseende centrerat kring att uppsampla lågupplösta bilder till högupplösta motsvarigheter. Detta examensarbete undersöker hur lämplig en specifik djupinlärningsmetod är för superupplösning i realtid. Enligt tidigare forskning använder vi oss av bikubisk interpolering som grund för jämförelse. Den valda djupinlärningsmetoden är speciellt anpassad till superupplösning i realtid och består av ett rörelsekompensationsnätverk och ett uppsamplingsnätverk. Djupainlärningsmetoden och interpoleringsmetoden jämförs genom att kvantitativt utvärdera metoderna mot varandra i kvalitetsmått och prestandamått. Lämpliga kvalitetsmått väljs från tidigare forskning för att ge ökad jämförbarhet mellan resultaten, nämligen maximalt signaltill- brusförhållande och strukturlikhetsindex. Prestandamätvärdena är: antal operationer för en uppsamplad bild, latens, genomströmning och minnesbehov. Vi utvärderar metoderna på ett utmanande allmänt tillgängligt dataset speciellt konstruerat för superupplösningsforskning inom video. För att ytterligare undersöka den djupa inlärningsmetoden föreslår vi några modifieringar och studerar effekten på mätvärdena. Våra resultat visar att djupinlärningsmodellerna överträffar bikubisk interpolering i kvalitetsmåtten, medan bikubisk interpolering överträffar djupinlärningsmodellerna i prestandamåtten. Vi finner inte heller någon signifikant kvalitetsmässig förbättring förknippad med att ha ett rörelsekompensationsnätverk för detta dataset, vilket kan betyda att datasetet är för komplext för rörelsekompensationnätverket. Vi drar slutsatsen att djupainlärningsmetoden uppvisar realtidsfunktioner eftersom metoden har en genomströmning på cirka 500 bilder per sekund för full HD superupplösning. Dessutom visar vi att genom att modifiera djupainlärningsmetoden uppnår vi liknande latens som bikubisk interpolering utan att offra genomströmning eller kvalitet.

Cette thèse porte sur le développement et l’évaluation de techniques d’imagerie en échocardiographie. L’objectif est de proposer des méthodes d’imagerie ultrasonore ultrarapide pour estimer le mouvement cardiaque 2-D et 3-D.Première modalité d’imagerie du cœur, l’échocardiographie conventionnelle permet la mesure des déformations myocardiques à 80 images/s. Cette cadence d’imagerie est insuffisante pour quantifier les mouvements de la totalité du myocarde lors de tests d’efforts, utiles en évaluation clinique, au cours desquels le rythme cardiaque est augmenté. De plus, la résolution temporelle actuelle en échocardiographie 3-D limite ses applications, pourtant essentielles pour une caractérisation complète du cœur.Les contributions présentées ici sont 1) le développement et l’évaluation, pour l’application cardiaque, d’une méthode originale d’estimation de mouvement 2-D par imagerie ultrarapide et marquage des images, 2) l’étude de faisabilité de la mesure globale des déformations cardiaques avec une méthode innovante d’imagerie ultrasonore ultrarapide 2-D et 3) la généralisation de cette approche en 3-D pour l’imagerie des volumes cardiaques à haute résolution temporelle. Cette technique est basée sur l’émission d’ondes divergentes, et l’intégration d’une compensation de mouvement dans le processus de formation des volumes cardiaques.La méthode proposée permet l’estimation des mouvements cardiaques 2-D et l’échocardiographie ultrarapide 3-D. L’évaluation de notre approche pour la quantification des déformations myocardiques locales 2-D et 3-D pourrait permettre de proposer des pistes innovantes pour poursuivre nos études et améliorer le diagnostic en routine clinique / This PhD work focuses on the development and the evaluation of imaging techniques in echocardiography. Our objective is to propose ultrafast ultrasound imaging methods for 2-D and 3-D cardiac motion estimations.Echocardiography is one of the most widespread modality for cardiovascular imaging. Conventional clinical scanners allow measurement of myocardial velocities and deformations at 80 images / s. In some situations, it can be recommended to increase the heart rate during a stress echocardiographic examination. Motion estimation of the whole myocardium at such heart rates is challenging with the conventional imaging systems. In addition, the low temporal resolution of the current conventional 3-D echocardiography limits quantitative applications, which would be needed for a complete characterization of the heart.The three contributions presented here are 1) the development and evaluation of an original method for 2-D cardiac motion estimation, with ultrafast imaging and image tagging, 2) the feasibility study of the global myocardial deformation measurement using an innovative 2-D ultrafast ultrasound imaging method and 3) the generalization of this approach in three dimensions for high frame-rate 3-D echocardiography. This method is based on the transmission of divergent waves and the integration of motion compensation, during the imaging process, to produce high-quality volumetric images of the heart.The proposed method allows 2-D cardiac motion estimation and 3-D echocardiography at high frame-rate. The evaluation of our approach for local 2-D and 3-D myocardial deformation measurements should permit to conduct further study in order to improve medical diagnosis

No description available.

Image processing algorithms have developed rapidly in recent years. Imaging functions are becoming more common in electronic devices, demanding better image quality, and more robust image capture in challenging conditions. Increasingly more complicated algorithms are being developed in order to achieve better signal to noise characteristics, more accurate colours, and wider dynamic range, in order to approach the human visual system performance levels.