In this work, a new method is proposed that can be used for area tracking. This method is based on the Complex Discrete Wavelet Transform (CDWT) developed by Magarey and Kingsbury. The CDWT has its advantages over the traditional Discrete Wavelet Transform such as approximate shift invariance, improved directional selectivity, and robustness to noise and illumination changes. The proposed method is a generalization of the CDWT based motion estimation method developed by Magarey and Kingsbury. The Complex Wavelet Tracker extends the original method to estimate the true motion of regions according to a parametric motion model. In this way, rotation, scaling, and shear type of motions can be handled in addition to pure translation. Simulations have been performed on the proposed method including both quantitative and qualitative tests. Quantitative tests are performed on synthetically created test sequences and results have been compared to true data. The performance is compared with intensity-based methods. Qualitative tests are performed on real sequences and evaluations are presented empirically. The results are compared with intensity-based methods. It is observed that the proposed method is very accurate in handling affine deformations for long term sequences and is robust to different target signatures and illumination changes. The accuracy of the proposed method is compatible with intensity-based methods. In addition to this, it can handle a wider range of cases and is robust to illuminaton changes compared to intensity-based methods. The method can be implemented in real-time and could be a powerful replacement of current area trackers.

The work presented in this thesis takes place in a context of growing demand for image and video applications on parallel embedded systems. The limitations and lack of flexibility of current design with parallel embedded systems make increasingly complicated to implement applications, particularly on heterogeneous systems. But Open Computing Language (OpenCL) is a new framework for fully employ the capability of computation of general purpose processors or embedded processors. In the meantime, some rapid prototyping tools to design systems are proposed to generate a reliably prototype or automatically implement the image and video applications on embedded systems. The goal of this thesis was to evaluate and to improve design processes for embedded systems, especially based on the dataflow approach (high level of abstraction) and OpenCL approach (intermediate level of abstraction). This challenge is tackled by several projects including the collaborative project COMPA which studies a framework based on the Orcc, Preesm and HMPP tools. In this context, this thesis aims to validate and to evaluate the framework with motion estimation and stereo matching algorithms. For this aim, algorithms have been described using the high-level RVC-CAL language. With the help of Orcc, Preesm, and HMPP tools, we generated and verified C code or OpenCL code or CUDA code for heterogeneous platforms based on multi-core CPU and GPU. We also studied the implementations of these algorithms onto the last generation of many-core for embedded system called MPPA and developed by KALRAY. We proposed three algorithms. One is a parallelized motion estimation method for heterogeneous system based on one CPU and one GPU: we developed one basic method to balance the workload distribution on such heterogeneous system. The second algorithm is a real-time stereo matching method that adopts combined costs and costs aggregation with square size step to implement on laptop's GPU platform: our experimental results outperform other baseline methods about tradeoff between matching accuracy and time-efficiency. The third algorithm is a joint motion-based video stereo matching method that uses the motion vectors calculated by the first algorithm to build the support region for the second algorithm: our experimental results outperform the stereo video matching methods in the test sequences with abundant movement even in large amounts of noise.

The objective of this thesis is to study the evolution of scalar field carried by a flow from a temporal image sequence. The estimation of the velocity field of turbulent flow is of major importance for understanding the physical phenomenon. Up to now the problem of turbulence is generally ignored in the flow equation of existing methods. An information given by image is discrete at pixel size. Depending on the turbulent rate of the flow, pixel and time resolutions may become too large to neglect the effect of sub-pixel small-scales on the pixel velocity field. For this, we propose a flow equation defined by a filtered concentration transport equation where a classic turbulent sub-grid eddy viscosity model is introduced in order to account for this effect. To formulate the problem, we use a Markovian approach. An unwarping multiresolution by pyramidal decomposition is proposed which reduces the number of operations on images. The optimization coupled with a multigrid approach allows to estimate the optimal 2D real velocity field. Our approach is tested on synthetic andreal image sequences (PIV laboratory experiment and remote sensing data of dust storm event) with high Reynolds number. Comparisons with existing approaches are very promising.

Made available in DSpace on 2014-12-17T15:47:47Z (GMT). No. of bitstreams: 1 MilanoGC.pdf: 520836 bytes, checksum: 93aa64380903b0b37d4a3f08b9900743 (MD5) Previous issue date: 2008-09-28 / Motion estimation is the main responsible for data reduction in digital video encoding. It is also the most computational damanding step. H.264 is the newest standard for video compression and was planned to double the compression ratio achievied by previous standards. It was developed by the ITU-T Video Coding Experts Group (VCEG) together with the ISO/IEC Moving Picture Experts Group (MPEG) as the product of a partnership effort known as the Joint Video Team (JVT). H.264 presents novelties that improve the motion estimation efficiency, such as the adoption of variable block-size, quarter pixel precision and multiple reference frames. This work defines an architecture for motion estimation in hardware/software, using a full search algorithm, variable block-size and mode decision. This work consider the use of reconfigurable devices, soft-processors and development tools for embedded systems such as Quartus II, SOPC Builder, Nios II and ModelSim / A estima??o de movimento ? a maior respons?vel pela redu??o na quantidade de dados na codifica??o de um v?deo digital, ? tamb?m a etapa que exige maior esfor?o computacional. O padr?o H.264 ? o mais novo padr?o de compress?o de v?deo e foi desenvolvido com o objetivo de dobrar a taxa de compress?o em rela??o aos demais padr?es existentes at? ent?o. Esse padr?o foi desenvolvido pelo JVT, que foi formado a partir de uma uni?o entre os especialistas do VCEG da ITU-T e do MPEG da ISO/IEC. O padr?o H.264 apresenta inova??es que aumentam a efici?ncia da estima??o de movimento, como o uso de blocos de tamanho vari?vel, precis?o de ? de pixel e a utiliza??o de m?ltiplos quadros de refer?ncia. A proposta deste trabalho ? a de definir uma arquitetura em hardware/software que realize a estima??o de movimento do padr?o H.264, utilizando o algoritmo da busca completa, blocos de tamanhos vari?veis e modo de decis?o. Para implementa??o este trabalho considera a utiliza??o de dispositivos reconfigur?veis, soft-processadores e ferramentas de desenvolvimento de sistemas embarcados tais como Quartus II, SOPC Builder, Nios II e ModelSim

Esta dissertação tem como objetivo estudar um método para estimar a velocidade de veículos ferroviários usando processamento de vídeo. O sistema proposto consiste na utilização de câmeras ao longo das vias férreas que permitam não só o monitoramento do tráfego ferroviário, mas cujo vídeo captado possa ser utilizado para a obtenção de estimativas para a velocidade instantânea dos trens que por ela trafegam. Tal sistema seria utilizado independentemente dos sistemas de controle já utilizados pela operadora do sistema ferroviário, permitindo que os controladores possam ter uma segunda análise no caso de falha da primeira, assim como um modelo que permita avaliar a velocidade instantânea do veículo ferroviário ao longo do percurso. Os algoritmos de rastreamento empregados para esse fim abordaram diferentes métodos. Confrontaram-se os resultados obtidos com os algoritmos propostos com os dados empíricos de forma a determinar aquele com melhor resposta dada as características do sistema. O algoritmo que apresentou os melhores resultados emprega um único bloco de referência para todos os quadros comparados. A métrica de similaridade responsável por determinar quais blocos são mais ou menos similares dentro do universo de busca estipulado é a soma de diferenças absolutas (SAD, Sum of Absolute Differences). O tempo de processamento requerido por cada um dos métodos de rastreamento estudados também foi considerado nas avaliações de resultados apresentadas. Uma comparação realizada entre as velocidades coletadas e aquelas informadas pelo sistema de controle mostraram que os resultados obtidos com o sistema atual, utilizando a sinalização apenas por circuito de via apresenta resultados pouco confiáveis com erros bastante significativos. Os resultados obtidos com o sistema proposto apresentaram erros menores quando comparados àqueles obtidos pelo sistema vigente, apresentando-se assim como uma solução viável e de baixo custo quando comparada às técnicas atualmente empregadas para a medida de velocidade de trens. / This dissertation aims to study a method of estimating the speed of railway vehicles using video processing. The proposed system uses cameras disposed alongside railways to monitor traffic as well as estimate train speeds. Such a system would allow the machinist to have a backup control system, regardless of the control systems already used by the operator of the rail system. In this dissertation, tracking algorithms with different methods are discussed. By comparing results obtained with the proposed algorithm and results from empirical data, it was possible to determine the one that best provided the desired results given the characteristics of the system. The algorithm which presented the best results was the one that employs a single block for all reference frames compared. The metric of similarity used for determining which blocks are most similar within the searched universe is the sum of absolute differences (SAD). The processing time required by each of the screening methods studied was also considered in assessing the results presented. A comparison between the rates collected and those informed by the control system showed that the results obtained with the current system, which only uses the signal track circuit, provides unreliable results with very significant errors. The results obtained with the proposed system had errors smaller than those obtained with the current system, suggesting the proposed system is viable and cost-effective compared to currently employed techniques.

Esta dissertação tem como objetivo estudar um método para estimar a velocidade de veículos ferroviários usando processamento de vídeo. O sistema proposto consiste na utilização de câmeras ao longo das vias férreas que permitam não só o monitoramento do tráfego ferroviário, mas cujo vídeo captado possa ser utilizado para a obtenção de estimativas para a velocidade instantânea dos trens que por ela trafegam. Tal sistema seria utilizado independentemente dos sistemas de controle já utilizados pela operadora do sistema ferroviário, permitindo que os controladores possam ter uma segunda análise no caso de falha da primeira, assim como um modelo que permita avaliar a velocidade instantânea do veículo ferroviário ao longo do percurso. Os algoritmos de rastreamento empregados para esse fim abordaram diferentes métodos. Confrontaram-se os resultados obtidos com os algoritmos propostos com os dados empíricos de forma a determinar aquele com melhor resposta dada as características do sistema. O algoritmo que apresentou os melhores resultados emprega um único bloco de referência para todos os quadros comparados. A métrica de similaridade responsável por determinar quais blocos são mais ou menos similares dentro do universo de busca estipulado é a soma de diferenças absolutas (SAD, Sum of Absolute Differences). O tempo de processamento requerido por cada um dos métodos de rastreamento estudados também foi considerado nas avaliações de resultados apresentadas. Uma comparação realizada entre as velocidades coletadas e aquelas informadas pelo sistema de controle mostraram que os resultados obtidos com o sistema atual, utilizando a sinalização apenas por circuito de via apresenta resultados pouco confiáveis com erros bastante significativos. Os resultados obtidos com o sistema proposto apresentaram erros menores quando comparados àqueles obtidos pelo sistema vigente, apresentando-se assim como uma solução viável e de baixo custo quando comparada às técnicas atualmente empregadas para a medida de velocidade de trens. / This dissertation aims to study a method of estimating the speed of railway vehicles using video processing. The proposed system uses cameras disposed alongside railways to monitor traffic as well as estimate train speeds. Such a system would allow the machinist to have a backup control system, regardless of the control systems already used by the operator of the rail system. In this dissertation, tracking algorithms with different methods are discussed. By comparing results obtained with the proposed algorithm and results from empirical data, it was possible to determine the one that best provided the desired results given the characteristics of the system. The algorithm which presented the best results was the one that employs a single block for all reference frames compared. The metric of similarity used for determining which blocks are most similar within the searched universe is the sum of absolute differences (SAD). The processing time required by each of the screening methods studied was also considered in assessing the results presented. A comparison between the rates collected and those informed by the control system showed that the results obtained with the current system, which only uses the signal track circuit, provides unreliable results with very significant errors. The results obtained with the proposed system had errors smaller than those obtained with the current system, suggesting the proposed system is viable and cost-effective compared to currently employed techniques.

Câmeras embarcadas em sistemas robóticos móveis com sensoriamento visual geralmente são afetadas pelo movimento de seu suporte quando é necessária a aquisição de imagens com alta confiabilidade. Algumas câmeras disponíveis no mercado já estão equipadas com sistemas de estabilização de imagens, implementados nas lentes da câmera ou no sensor de imagem. Esses sistemas são relativamente caros para serem instalados em sistemas robóticos móveis aéreos ou terrestres de pequeno porte. Outro conceito que tem sido utilizado para alcançar a estabilização consiste em adotar sensores inerciais, algoritmos de visão computacional e aplicação de filtros digitais para estimativa e suavização de movimento. Esta dissertação tem por finalidade apresentar um sistema de estabilização digital de vídeo em conjunto com um acelerômetro para detectar movimentos da câmera. Considera-se que a câmera está instalada sobre plataforma instável ou em veículo em movimento, assim, utiliza-se filtragem robusta para minimizar o efeito da vibração sobre a câmera. Vários experimentos são realizados adicionando vibração em vídeo por meio de simulações em computador. Considera-se também experimentos em ambiente real com a câmera montada em um suporte dentro de um veículo em movimento. A principal contribuição deste trabalho é a realização de um estudo comparativo para analisar a vantagem de se usar uma abordagem robusta entre os métodos de filtragem comumente utilizados em sistemas de estabilização de vídeo. / Embedded cameras in robotic systems usually are susceptible to movements of your basis. The measurements can be affected mainly when they should obtain high performance in actual applications. Although some cameras available on the market are already equipped with an optical image stabilization system, implemented either in the camera lenses or in the image sensor. They are usually expensive to be installed into small aerial or land robotic systems. This dissertation presents a video stabilization system coupled with an accelerometer for motion detection at an unstable platform. It is embedded in a vehicle. Robust filtering is used to minimize vibration effects on the camera. Several experiments were performed, adding either vibrations via computer simulations and taking images from the camera with the vehicle moving. The main contribution of this work is the accomplishment of a comparative study to analyse the advantages of applying a robust approach among others filtering methods that frequently are used to video stabilizing systems.

L’objectif de cette thèse est l’étude de l’évolution de champ scalaire transporté par un écoulement à partir d’une séquence d’images temporelles. L’estimation du champ de vitesse d’un écoulement turbulent est d’une importance majeure pour mieux comprendre le phénomène physique. Jusqu’à présent, le problème de la turbulence est généralement ignoré dans l’équation de mouvement des méthodes existantes. Les images contiennent une information discrète correspondant à la taille du pixel. Selon le niveau de turbulence de l’écoulement, les résolutions des pixels et du temps peuvent devenir trop grandes pour négliger l’effet des petites échelles (sous-pixel) sur le champ de vitesse. Nous proposons pour cela, une équation de mouvement définie par l’équation de transport de concentration filtrée pour laquelle un modèle classique de viscosité turbulente sous-maille est introduit afin de tenir compte de cet effet. Nous utilisons pour formuler le problème, une approche Markovienne. Une méthode de multirésolution par décomposition pyramidale, sans transformation d’image intermédiaire au cours du processus, est proposée. Cela permet de diminuer le nombre d’opérations sur les images. La méthode d’optimisation utilisée, couplée avec une approche multigrille, permet d’obtenir le champ de vitesse réel optimal. Notre approche est testée sur des séquences d’images synthétiques et réelles (expérience PIV et tempête de sable à partir d’image de télédétection) avec des nombres de Reynolds élevés. Les comparaisons avec des approches existantes sont très prometteuses. / The objective of this thesis is to study the evolution of scalar field carried by a flow from a temporal image sequence. The estimation of the velocity field of turbulent flow is of major importance for understanding the physical phenomenon. Up to now the problem of turbulence is generally ignored in the flow equation of existing methods. An information given by image is discrete at pixel size. Depending on the turbulent rate of the flow, pixel and time resolutions may become too large to neglect the effect of sub-pixel small-scales on the pixel velocity field. For this, we propose a flow equation defined by a filtered concentration transport equation where a classic turbulent sub-grid eddy viscosity model is introduced in order to account for this effect. To formulate the problem, we use a Markovian approach. An unwarping multiresolution by pyramidal decomposition is proposed which reduces the number of operations on images. The optimization coupled with a multigrid approach allows to estimate the optimal 2D real velocity field. Our approach is tested on synthetic andreal image sequences (PIV laboratory experiment and remote sensing data of dust storm event) with high Reynolds number. Comparisons with existing approaches are very promising.

Cette étude a pour cadre l’analyse d’images dans un contexte multi-échelles, une attention particulière étant portée sur les images fluides dans un contexte turbulent. Nous traitons en premier lieu le problème de l’estimation de mouvement. Dans un contexte multi-échelles, on néglige bien souvent dans un premier temps la contribution des fines échelles du problème. Nous proposons, pour pallier ce problème, plusieurs termes d’attache aux données dérivant de l’OFCE. Ceux-ci permettent, à chaque niveau d’échelle, la prise en compte de ces composantes fines échelles. Les performances de ces termes sont expérimentalement démontrées sur des images générales et fluides. Nous abordons en second lieu le problème de super-résolution d’images de scalaires passifs : nous souhaitons reconstruire, de manière explicite, certains détails manquants au sein d’images basse résolution données. Pour cela, nous utilisons plusieurs modèles issus de la simulation des grandes échelles ainsi que des méthodes d’assimilation de données permettant d’assurer une certaine cohérence temporelle de la solution. Les approches présentées sont expérimentalement étudiées sur différentes séquences d’images. Enfin, nous proposons une méthode d’estimation multi-résolution permettant de combiner de manière simultanée les informations issues des différents niveaux de résolution. / This thesis is concerned with image analysis within a multi-scale framework. Specific attention is given to fluid images in the presence of turbulence. In a first part, we adress the problem of multi-scale motion estimation from image sequences. Starting from OFCE equation, we derive several image data terms allowing to take into account, while estimating the solution coarse scales, the contribution of the finer scales components usually neglected in classic approaches. The performances of the resulting estimators is demonstrated on both general and fluid images. The second step of this study is concerned with the problem of passive scalar images super- resolution : starting from low resolution input images, we aim at recovering some of the missing high frequencies. We proposed several methods inspired by the LES framework, as well as data assimilation techniques, in order to ensure the solution consistency over time. Such approaches are experimented and compared over various image sequences. Finally, we propose a multi-resolution estimation method simultaneously combining informations from different grid levels.

Nous nous intéressons dans cette thèse au problème de l'estimation dense du mouvement dans des séquences d'images, également désigné sous le terme de flot optique. Les approches usuelles exploitent une paramétrisation locale ou une régularisation globale du champ de déplacement. Nous explorons plusieurs façons de combiner ces deux stratégies, pour surmonter leurs limitations respectives. Nous nous plaçons dans un premier temps dans un cadre variationnel global, et considérons un filtrage local du terme de données. Nous proposons un filtrage spatialement adaptatif, optimisé conjointement au mouvement, pour empêcher le sur-lissage induit par le filtrage spatialement constant. Dans une seconde partie, nous proposons un cadre générique d'agrégation pour l'estimation du flot optique. Sous sa forme générale, il consiste en une estimation locale de candidats de mouvements, suivie de leur combinaison à l'étape d'agrégation avec un modèle global. Ce schéma permet une estimation efficace des grands déplacements et des discontinuités de mouvement. Nous développons également une méthode générique de gestion des occultations. Notre méthode est validée par une analyse expérimentale conséquente sur des bases de données de référence en vision par ordinateur. Nous démontrons la supériorité de notre méthode par rapport à l'état de l'art sur les séquences présentant de grands déplacements. La dernière partie de la thèse est consacrée à l'adaptation des approches précédentes à des problématiques d'imagerie biologique. Les changements locaux importants d'intensité observés en imagerie de fluorescence sont estimés et compensé par une adaptation de notre schéma d'agrégation. Nous proposons également une méthode variationnelle avec filtrage local dédiée au cas de mouvements diffusifs de particules. / This thesis is concerned with dense motion estimation in image sequences, also known as optical flow. Usual approaches exploit either local parametrization or global regularization of the motion field. We explore several ways to combine these two strategies, to overcome their respective limitations. We first address the problem in a global variational framework, and consider local filtering of the data term. We design a spatially adaptive filtering optimized jointly with motion, to prevent over-smoothing induced by the spatially constant approach. In a second part, we propose a generic two-step aggregation framework for optical flow estimation. The most general form is a local computation of motion candidates, combined in the aggregation step through a global model. Large displacements and motion discontinuities are efficiently recovered with this scheme. We also develop a generic exemplar-based occlusion handling to deal with large displacements. Our method is validated with extensive experiments in computer vision benchmarks. We demonstrate the superiority of our method over state-of-the-art on sequences with large displacements. Finally, we adapt the previous methods to biological imaging issues. Estimation and compensation of large local intensity changes frequently occurring in fluorescence imaging are efficiently estimated and compensated with an adaptation of our aggregation framework. We also propose a variational method with local filtering dedicated to the case of diffusive motion of particles.