Les fondements théoriques pour la saillance visuelle ont été dressés, il y a 35 ans, par Treisman qui a proposé "feature-integration theory" pour le système visuel humain: dans n’importe quel contenu visuel, certaines régions sont saillantes en raison de la différence entre leurs caractéristiques (intensité, couleur, texture, et mouvement) et leur voisinage. Notre thèse offre un cadre méthodologique et expérimental compréhensif pour extraire les régions saillantes directement des flux compressés (MPEG-4 AVC et HEVC), tout en minimisant les opérations de décodage. L’extraction de la saillance visuelle à partir du flux compressé est à priori une contradiction conceptuelle. D’une part, comme suggéré par Treisman, dans un contenu vidéo, la saillance est donnée par des singularités visuelles. D’autre part, afin d’éliminer la redondance visuelle, les flux compressés ne devraient plus préserver des singularités. La thèse souligne également l’avantage pratique de l’extraction de la saillance dans le domaine compressé. Dans ce cas, nous avons démontré que, intégrée dans une application de tatouage robuste de la vidéo compressée, la carte saillance agit comme un outil d’optimisation, ce qui permet d’augmenter la transparence (pour une quantité d’informations insérées et une robustesse contre les attaques prescrites) tout en diminuant la complexité globale du calcul. On peut conclure que la thèse démontre aussi bien méthodologiquement que expérimentalement que même si les normes MPEG-4 AVC et HEVC ne dépendent pas explicitement d’aucun principe de saillance visuelle, leurs flux préservent cette propriété remarquable reliant la représentation numérique de la vidéo au mécanisme psycho-cognitifs humains / The theoretical ground for visual saliency was established some 35 years ago by Treisman who advanced the integration theory for the human visual system: in any visual content, some regions are salient (appealing) because of the discrepancy between their features (intensity, color, texture, motion) and the features of their surrounding areas. This present thesis offers a comprehensive methodological and experimental framework for extracting the salient regions directly from video compressed streams (namely MPEG-4 AVC and HEVC), with minimal decoding operations. Note that saliency extraction from compressed domain is a priori a conceptual contradiction. On the one hand, as suggested by Treisman, saliency is given by visual singularities in the video content. On the other hand, in order to eliminate the visual redundancy, the compressed streams are no longer expected to feature singularities. The thesis also brings to light the practical benefit of the compressed domain saliency extraction. In this respect, the case of robust video watermarking is targeted and it is demonstrated that the saliency acts as an optimization tool, allowing the transparency to be increased (for prescribed quantity of inserted information and robustness against attacks) while decreasing the overall computational complexity. As an overall conclusion, the thesis methodologically and experimentally demonstrates that although the MPEG-4 AVC and the HEVC standards do not explicitly rely on any visual saliency principle, their stream syntax elements preserve this remarkable property linking the digital representation of the video to sophisticated psycho-cognitive mechanisms

The human vision system is generally recognized as being superior to all known artificial vision systems. Visual attention, among many processes that are related to human vision, is responsible for identifying relevant regions in a scene for further processing. In most cases, analyzing an entire scene is unnecessary and inevitably time consuming. Hence considering visual attention might be advantageous. A subfield of computer vision where this particular functionality is computationally emulated has been shown to retain high potential in solving real world vision problems effectively. In this monograph, elements of visual attention are explored and algorithms are proposed that exploit such elements in order to enhance image understanding capabilities. Satellite images are given special attention due to their practical relevance, inherent complexity in terms of image contents, and their resolution. Processing such large-size images using visual attention can be very helpful since one can first identify relevant regions and deploy further detailed analysis in those regions only. Bottom-up features, which are directly derived from the scene contents, are at the core of visual attention and help identify salient image regions. In the literature, the use of intensity, orientation and color as dominant features to compute bottom-up attention is ubiquitous. The effects of incorporating an entropy feature on top of the above mentioned ones are also studied. This investigation demonstrates that such integration makes visual attention more sensitive to fine details and hence retains the potential to be exploited in a suitable context. One interesting application of bottom-up attention, which is also examined in this work, is that of image segmentation. Since low salient regions generally correspond to homogenously textured regions in the input image; a model can therefore be learned from a homogenous region and used to group similar textures existing in other image regions. Experimentation demonstrates that the proposed method produces realistic segmentation on satellite images. Top-down attention, on the other hand, is influenced by the observer’s current states such as knowledge, goal, and expectation. It can be exploited to locate target objects depending on various features, and increases search or recognition efficiency by concentrating on the relevant image regions only. This technique is very helpful in processing large images such as satellite images. A novel algorithm for computing top-down attention is proposed which is able to learn and quantify important bottom-up features from a set of training images and enhances such features in a test image in order to localize objects having similar features. An object recognition technique is then deployed that extracts potential target objects from the computed top-down attention map and attempts to recognize them. An object descriptor is formed based on physical appearance and uses both texture and shape information. This combination is shown to be especially useful in the object recognition phase. The proposed texture descriptor is based on Legendre moments computed on local binary patterns, while shape is described using Hu moment invariants. Several tools and techniques such as different types of moments of functions, and combinations of different measures have been applied for the purpose of experimentations. The developed algorithms are generalized, efficient and effective, and have the potential to be deployed for real world problems. A dedicated software testing platform has been designed to facilitate the manipulation of satellite images and support a modular and flexible implementation of computational methods, including various components of visual attention models.

The human vision system is generally recognized as being superior to all known artificial vision systems. Visual attention, among many processes that are related to human vision, is responsible for identifying relevant regions in a scene for further processing. In most cases, analyzing an entire scene is unnecessary and inevitably time consuming. Hence considering visual attention might be advantageous. A subfield of computer vision where this particular functionality is computationally emulated has been shown to retain high potential in solving real world vision problems effectively. In this monograph, elements of visual attention are explored and algorithms are proposed that exploit such elements in order to enhance image understanding capabilities. Satellite images are given special attention due to their practical relevance, inherent complexity in terms of image contents, and their resolution. Processing such large-size images using visual attention can be very helpful since one can first identify relevant regions and deploy further detailed analysis in those regions only. Bottom-up features, which are directly derived from the scene contents, are at the core of visual attention and help identify salient image regions. In the literature, the use of intensity, orientation and color as dominant features to compute bottom-up attention is ubiquitous. The effects of incorporating an entropy feature on top of the above mentioned ones are also studied. This investigation demonstrates that such integration makes visual attention more sensitive to fine details and hence retains the potential to be exploited in a suitable context. One interesting application of bottom-up attention, which is also examined in this work, is that of image segmentation. Since low salient regions generally correspond to homogenously textured regions in the input image; a model can therefore be learned from a homogenous region and used to group similar textures existing in other image regions. Experimentation demonstrates that the proposed method produces realistic segmentation on satellite images. Top-down attention, on the other hand, is influenced by the observer’s current states such as knowledge, goal, and expectation. It can be exploited to locate target objects depending on various features, and increases search or recognition efficiency by concentrating on the relevant image regions only. This technique is very helpful in processing large images such as satellite images. A novel algorithm for computing top-down attention is proposed which is able to learn and quantify important bottom-up features from a set of training images and enhances such features in a test image in order to localize objects having similar features. An object recognition technique is then deployed that extracts potential target objects from the computed top-down attention map and attempts to recognize them. An object descriptor is formed based on physical appearance and uses both texture and shape information. This combination is shown to be especially useful in the object recognition phase. The proposed texture descriptor is based on Legendre moments computed on local binary patterns, while shape is described using Hu moment invariants. Several tools and techniques such as different types of moments of functions, and combinations of different measures have been applied for the purpose of experimentations. The developed algorithms are generalized, efficient and effective, and have the potential to be deployed for real world problems. A dedicated software testing platform has been designed to facilitate the manipulation of satellite images and support a modular and flexible implementation of computational methods, including various components of visual attention models.

Thesis (MScEng)--Stellenbosch University, 2014. / ENGLISH ABSTRACT: The focus of this thesis is on the optimal design and evaluation of FI-PM machines to be used with a MG transmission drive-train for EV application. The machines presented are optimised using a gradient-based optimisation algorithm of the VisualDoc software together with FE software and Python scripts. Each machine is optimised for its own objective function. The focus is to reduce expensive rare earth material. High torque ripple issues of the optimised machines are solved by implementing a relatively new topology where the rotor poles/barriers are made asymmetric. The asymmetric rotor topology implemented is effective and can be used as an alternative for rotor stack skewing. PM demagnetisation and rotor deformation studies are conducted on the optimum designed machines to ensure that no PM demagnetisation on the surface of the PMs and critical rotor deformation occur. The FE performance results of the optimum designed machines are shown and discussed. One of the optimum designed FI-PM machines is manufactured and tested in the laboratory. The FE and measured results of the machine are compared and shows good correlation. The saliency performance of the optimum designed machines are evaluated as it determines its position sensorless control capability. It is shown that the saliency ratios increase linearly with load, making it favourable for position sensorless control. It is also shown that the asymmetric rotor topologies introduced a larger mutual inductance compared to their symmetric counterparts, thus higher cross-coupling is present in these rotors and therefore a higher saliency shift, which is undesirable. Two case studies are performed in order to improve saliency performance. The objective of the first case study is to improve the saliency shift by reducing the flux leakage paths in the rotor. The objective of the second case study is to optimise a FI-PM machine in order to improve the saliency ratio and -shift. The results of the two case studies are compared with the saliency performance of the other machines. / AFRIKAANSE OPSOMMING: Die fokus van hierdie tesis is op die optimale ontwerp en evaluering van veld versterking permanente magneet masjiene vir veelvoudige-rat elektriese voertuig toepassings. Die masjiene teenwoordig is geoptimeer met behulp van ’n helling-gebaseerde optimering algoritme. Elke masjien is geoptimeer vir sy eie doel funksie. Die fokus is om duur seldsame permanent magneet materiaal te verminder. Hoë wringkrag-rimpeleffek van die optimale masjiene word opgelos deur die implementering van ’n relatief nuwe topologie waar die rotor pole/vloedbarrière asimmetries gemaak word. Die asimmetriese rotor topologie wat geimplementeer is, is effektief en kan dus as ’n alternatief vir die rotor stapel skeef metode gebruik word. Permanent magneet demagnetisering en rotor vervorming studies is ook uitgevoer op die optimum ontwerpte masjiene om te verseker dat geen demagnetisering plaasvind nie en ook geen kritiese rotor vervorming nie. Die eindige-element resultate van die optimum ontwerpte masjiene word getoon en bespreek. Een van die optimum ontwerpte veld versterking permanente magneet masjiene is vervaardig en getoets in die laboratorium . Die eindige-element en gemete resultate van die masjien word vergelyk en toon goeie korrelasie. Die speek prestasie van die optimum ontwerpte masjiene word geëvalueer aangesien dit die sensorlose posisie beheer vermoë bepaal. Daar word getoon dat die speek koëffisiënt verhoog lineêr met vrag wat dit gunstig maak vir posisie sensorlose beheer . Daar word ook gewys dat die asimmetriese rotor topologie ’n groter wedersydse induktansie het in vergelyking met hul simmetriese eweknieë, dus is daar hoër kruis-koppeling teenwoordig in die rotors en dus ’n ho¨er speek skuif, wat ongewens is. Twee gevallestudies om speek prestasie te verbeter is uitgevoer. Die doel van die eerste gevallestudie is om die speek skuif te verbeter deur die vermindering van die vloed lekkasie paaie in die rotor. Die doel van die tweede gevallestudie is om ’n veld versterking permanente magneet masjiene te optimeer ten einde die speek koëffisiënt en - skuif te verbeter. Die resultate van die twee gevallestudies word vergelyk met die speek prestasie van die ander masjiene.

Considerações de informação perceptual têm ganhado espaço rapidamente em pesquisas referentes a representação, análise e exibição de malhas. Estudos com usuários, eye tracking e outras técnicas são capazes de fornecer informações cada vez mais úteis para sistemas voltados a usuário, que formam a maioria das aplicações em computação gráfica. Neste trabalho nós expandimos sobre o conceito de Saliência de Malhas — uma medida automática de importância visual para malhas de triângulos baseada em modelos de atenção humana em baixo nível — melhorando, extendendo e realizando integração com diferentes aplicações. Nós extendemos o conceito de Saliência de Malhas para englobar objetos deformáveis, mostrando como um mapa de saliência em nível de vértice pode ser construído capturando corretamente regiões de alta importância perceptual através de um conjunto de poses ou deformações. Nós definimos saliência multi-pose como um agregado multi-escala de valores de curvatura sobre uma vizinhança localmente estável, em conjunto com deformações desta vizinhança em múltiplas poses. Nós substituímos distância Euclideana por geodésica, assim fornecendo melhores estimativas de vizinhança local. Resultados mostram que saliência multi-pose gera resultados visualmente mais interessantes em simplificações quando comparado à saliência em uma única pose. Nós também aplicamos saliência de malhas ao problema de segmentação e rendering dependente de ponto de vista, introduzindo uma técnica para segmentação que particiona um objeto em um conjunto de clusters, cada um englobando um grupo de características localmente interessantes. Saliência de malhas é incorporada em um framework para clustering propagativo, guiando seleção de pontos de partida para clusters e custos de propagação de faces, levando a uma convergência de clusters ao redor de características perceptualmente importantes. Nós comparamos nossa técnica com diferentes métodos automáticos para segmentação, mostrando que ela fornece segmentação melhor ou comparável sem necessidade de intervenção do usuário. Uma vez que o algoritmo de segmentação proposto é especialmente aplicável a rendering multi-resolução, nós ilustramos uma aplicação do mesmo através de um sistema de rendering baseado em ponto de vista guiado por saliência, alcançando melhorias consideráveis em framerate com muito pouca perda de qualidade visual. / Considerations on perceptual information are quickly gaining importance in mesh representation, analysis and display research. User studies, eye tracking and other techniques are able to provide ever more useful insights for many user-centric systems, which form the bulk of computer graphics applications. In this work we build upon the concept of Mesh Saliency — an automatic measure of visual importance for triangle meshes based on models of low-level human visual attention—improving, extending and integrating it with different applications. We extend the concept of Mesh Saliency to encompass deformable objects, showing how a vertex-level saliency map can be constructed that accurately captures the regions of high perceptual importance over a range of mesh poses or deformations. We define multipose saliency as a multi-scale aggregate of curvature values over a locally stable vertex neighborhood together with deformations over multiple poses. We replace the use of the Euclidean distance by geodesic distance thereby providing superior estimates of the local neighborhood. Results show that multi-pose saliency generates more visually appealing mesh simplifications when compared to a single-pose mesh saliency. We also apply Mesh Saliency to the problem of mesh segmentation and view-dependent rendering, introducing a technique for segmentation that partitions an object into a set of face clusters, each encompassing a group of locally interesting features. Mesh Saliency is incorporated in a propagative mesh clustering framework, guiding cluster seed selection and triangle propagation costs and leading to a convergence of face clusters around perceptually important features. We compare our technique with different fully automatic segmentation algorithms, showing that it provides similar or better segmentation without the need for user input. Since the proposed clustering algorithm is specially suitable for multi-resolution rendering, we illustrate application of our clustering results through a saliency-guided view-dependent rendering system, achieving significant framerate increases with little loss of visual detail.

Considerações de informação perceptual têm ganhado espaço rapidamente em pesquisas referentes a representação, análise e exibição de malhas. Estudos com usuários, eye tracking e outras técnicas são capazes de fornecer informações cada vez mais úteis para sistemas voltados a usuário, que formam a maioria das aplicações em computação gráfica. Neste trabalho nós expandimos sobre o conceito de Saliência de Malhas — uma medida automática de importância visual para malhas de triângulos baseada em modelos de atenção humana em baixo nível — melhorando, extendendo e realizando integração com diferentes aplicações. Nós extendemos o conceito de Saliência de Malhas para englobar objetos deformáveis, mostrando como um mapa de saliência em nível de vértice pode ser construído capturando corretamente regiões de alta importância perceptual através de um conjunto de poses ou deformações. Nós definimos saliência multi-pose como um agregado multi-escala de valores de curvatura sobre uma vizinhança localmente estável, em conjunto com deformações desta vizinhança em múltiplas poses. Nós substituímos distância Euclideana por geodésica, assim fornecendo melhores estimativas de vizinhança local. Resultados mostram que saliência multi-pose gera resultados visualmente mais interessantes em simplificações quando comparado à saliência em uma única pose. Nós também aplicamos saliência de malhas ao problema de segmentação e rendering dependente de ponto de vista, introduzindo uma técnica para segmentação que particiona um objeto em um conjunto de clusters, cada um englobando um grupo de características localmente interessantes. Saliência de malhas é incorporada em um framework para clustering propagativo, guiando seleção de pontos de partida para clusters e custos de propagação de faces, levando a uma convergência de clusters ao redor de características perceptualmente importantes. Nós comparamos nossa técnica com diferentes métodos automáticos para segmentação, mostrando que ela fornece segmentação melhor ou comparável sem necessidade de intervenção do usuário. Uma vez que o algoritmo de segmentação proposto é especialmente aplicável a rendering multi-resolução, nós ilustramos uma aplicação do mesmo através de um sistema de rendering baseado em ponto de vista guiado por saliência, alcançando melhorias consideráveis em framerate com muito pouca perda de qualidade visual. / Considerations on perceptual information are quickly gaining importance in mesh representation, analysis and display research. User studies, eye tracking and other techniques are able to provide ever more useful insights for many user-centric systems, which form the bulk of computer graphics applications. In this work we build upon the concept of Mesh Saliency — an automatic measure of visual importance for triangle meshes based on models of low-level human visual attention—improving, extending and integrating it with different applications. We extend the concept of Mesh Saliency to encompass deformable objects, showing how a vertex-level saliency map can be constructed that accurately captures the regions of high perceptual importance over a range of mesh poses or deformations. We define multipose saliency as a multi-scale aggregate of curvature values over a locally stable vertex neighborhood together with deformations over multiple poses. We replace the use of the Euclidean distance by geodesic distance thereby providing superior estimates of the local neighborhood. Results show that multi-pose saliency generates more visually appealing mesh simplifications when compared to a single-pose mesh saliency. We also apply Mesh Saliency to the problem of mesh segmentation and view-dependent rendering, introducing a technique for segmentation that partitions an object into a set of face clusters, each encompassing a group of locally interesting features. Mesh Saliency is incorporated in a propagative mesh clustering framework, guiding cluster seed selection and triangle propagation costs and leading to a convergence of face clusters around perceptually important features. We compare our technique with different fully automatic segmentation algorithms, showing that it provides similar or better segmentation without the need for user input. Since the proposed clustering algorithm is specially suitable for multi-resolution rendering, we illustrate application of our clustering results through a saliency-guided view-dependent rendering system, achieving significant framerate increases with little loss of visual detail.

Nous étudions l'influence de différents attributs audiovisuels sur l'exploration visuelle de scènes naturelles dynamiques. Nous démontrons que si la façon dont nous explorons une scène dépend avant tout de son contenu visuel, dans certaines situations le son influence significativement les mouvements oculaires. La présence de son assure une meilleure cohérence entre les positions oculaires de différents observateurs, attirant leur attention et donc leur regard vers les mêmes régions. L'effet du son se retrouve tout particulièrement dans les scènes de conversation, où la présence du signal de parole associé augmente le nombre de fixations sur le visage des locuteurs, et donc la cohérence entre les scanpaths. Nous proposons un modèle de saillance audiovisuelle repérant automatiquement le visage des locuteurs afin d'en rehausser la saillance. Ces résultats s'appuient sur les mouvements oculaires de 148 participants enregistrés sur un total de plus de 75 400 frames (125 vidéos) dans 5 conditions expérimentales différentes. / We study the influence of different audiovisual features on the visualexploration of dynamic natural scenes. We show that, whilst the way a person explores a scene primarily relies on its visual content, sound sometimes significantly influences eye movements. Sound assures a better coherence between the eye positions of different observers, attracting their attention and thus their gaze toward the same regions. The effect of sound is particularly strong in conversation scenes, where the related speech signal boosts the number of fixations on speakers' faces, and thus increases the consistency between scanpaths. We propose an audiovisual saliency model able to automatically locate speakers' faces so as to enhance their saliency. These results are based on the eye movements of 148 participants recorded on more than 75,400 frames (125 videos) in 5 different experimental conditions.

Considerações de informação perceptual têm ganhado espaço rapidamente em pesquisas referentes a representação, análise e exibição de malhas. Estudos com usuários, eye tracking e outras técnicas são capazes de fornecer informações cada vez mais úteis para sistemas voltados a usuário, que formam a maioria das aplicações em computação gráfica. Neste trabalho nós expandimos sobre o conceito de Saliência de Malhas — uma medida automática de importância visual para malhas de triângulos baseada em modelos de atenção humana em baixo nível — melhorando, extendendo e realizando integração com diferentes aplicações. Nós extendemos o conceito de Saliência de Malhas para englobar objetos deformáveis, mostrando como um mapa de saliência em nível de vértice pode ser construído capturando corretamente regiões de alta importância perceptual através de um conjunto de poses ou deformações. Nós definimos saliência multi-pose como um agregado multi-escala de valores de curvatura sobre uma vizinhança localmente estável, em conjunto com deformações desta vizinhança em múltiplas poses. Nós substituímos distância Euclideana por geodésica, assim fornecendo melhores estimativas de vizinhança local. Resultados mostram que saliência multi-pose gera resultados visualmente mais interessantes em simplificações quando comparado à saliência em uma única pose. Nós também aplicamos saliência de malhas ao problema de segmentação e rendering dependente de ponto de vista, introduzindo uma técnica para segmentação que particiona um objeto em um conjunto de clusters, cada um englobando um grupo de características localmente interessantes. Saliência de malhas é incorporada em um framework para clustering propagativo, guiando seleção de pontos de partida para clusters e custos de propagação de faces, levando a uma convergência de clusters ao redor de características perceptualmente importantes. Nós comparamos nossa técnica com diferentes métodos automáticos para segmentação, mostrando que ela fornece segmentação melhor ou comparável sem necessidade de intervenção do usuário. Uma vez que o algoritmo de segmentação proposto é especialmente aplicável a rendering multi-resolução, nós ilustramos uma aplicação do mesmo através de um sistema de rendering baseado em ponto de vista guiado por saliência, alcançando melhorias consideráveis em framerate com muito pouca perda de qualidade visual. / Considerations on perceptual information are quickly gaining importance in mesh representation, analysis and display research. User studies, eye tracking and other techniques are able to provide ever more useful insights for many user-centric systems, which form the bulk of computer graphics applications. In this work we build upon the concept of Mesh Saliency — an automatic measure of visual importance for triangle meshes based on models of low-level human visual attention—improving, extending and integrating it with different applications. We extend the concept of Mesh Saliency to encompass deformable objects, showing how a vertex-level saliency map can be constructed that accurately captures the regions of high perceptual importance over a range of mesh poses or deformations. We define multipose saliency as a multi-scale aggregate of curvature values over a locally stable vertex neighborhood together with deformations over multiple poses. We replace the use of the Euclidean distance by geodesic distance thereby providing superior estimates of the local neighborhood. Results show that multi-pose saliency generates more visually appealing mesh simplifications when compared to a single-pose mesh saliency. We also apply Mesh Saliency to the problem of mesh segmentation and view-dependent rendering, introducing a technique for segmentation that partitions an object into a set of face clusters, each encompassing a group of locally interesting features. Mesh Saliency is incorporated in a propagative mesh clustering framework, guiding cluster seed selection and triangle propagation costs and leading to a convergence of face clusters around perceptually important features. We compare our technique with different fully automatic segmentation algorithms, showing that it provides similar or better segmentation without the need for user input. Since the proposed clustering algorithm is specially suitable for multi-resolution rendering, we illustrate application of our clustering results through a saliency-guided view-dependent rendering system, achieving significant framerate increases with little loss of visual detail.

De nombreuses applications de la vision par ordinateur requièrent la détection, la localisation et le suivi de régions ou d’objets d’intérêt dans une image ou une séquence d’images. De nombreux modèles d’attention visuelle, inspirés de la vision humaine, qui détectent de manière automatique les régions d’intérêt dans une image ou une vidéo, ont récemment été développés et utilisés avec succès dans différentes applications. Néanmoins, la plupart des approches existantes sont limitées à l’analyse de scènes statiques et très peu de méthodes exploitent la nature temporelle des séquences d’images.L'objectif principal de ce travail de thèse est donc l'étude de modèles d'attention visuelle pour l'analyse de scènes dynamiques complexes. Une carte de saliance est habituellement obtenue par la fusion d'une carte statitque (saliance spatiale dans une image) d'une part, et d'une carte dynamique (salience temporelle entre une série d'image) d'autre part. Dans notre travail, nous modélisons les changements dynamiques par un opérateur de texture LBP-TOP (Local Binary Patterns) et nous utilisons l'information couleur pour l'aspect spatial.Les deux cartes de saliances sont calculées en utilisant une formulation discriminante inspirée du système visuel humain, et fuionnées de manière appropriée en une carte de saliance spatio-temporelle.De nombreuses expériences avec des bases de données publiques, montrent que notre approche obteint des résulats meilleurs ou comparables avec les approches de la littérature. / Visual saliency is an important research topic in the field of computer vision due to its numerouspossible applications. It helps to focus on regions of interest instead of processingthe whole image or video data. Detecting visual saliency in still images has been widelyaddressed in literature with several formulations. However, visual saliency detection invideos has attracted little attention, and is a more challenging task due to additional temporalinformation. Indeed, a video contains strong spatio-temporal correlation betweenthe regions of consecutive frames, and, furthermore, motion of foreground objects dramaticallychanges the importance of the objects in a scene. The main objective of thethesis is to develop a spatio-temporal saliency method that works well for complex dynamicscenes.A spatio-temporal saliency map is usually obtained by the fusion of a static saliency mapand a dynamic saliency map. In our work, we model the dynamic textures in a dynamicscene with Local Binary Patterns (LBP-TOP) to compute the dynamic saliency map, andwe use color features to compute the static saliency map. Both saliency maps are computedusing a bio-inspired mechanism of Human Visual System (HVS) with a discriminantformulation known as center surround saliency, and are fused in a proper way.The proposed models have been extensively evaluated with diverse publicly availabledatasets which contain several videos of dynamic scenes. The evaluation is performed intwo parts. First, the method in locating interesting foreground objects in complex scene.Secondly, we evaluate our model on the task of predicting human observers fixations.The proposed method is also compared against state-of-the art methods, and the resultsshow that the proposed approach achieves competitive results.In this thesis we also evaluate the performance of different fusion techniques, because fusionplays a critical role in the accuracy of the spatio-temporal saliency map. We evaluatethe performances of different fusion techniques on a large and diverse complex datasetand the results show that a fusion method must be selected depending on the characteristics,in terms of color and motion contrasts, of a sequence. Overall, fusion techniqueswhich take the best of each saliency map (static and dynamic) in the final spatio-temporalmap achieve best results.

Le contrôle aérien rentre dans une phase de profondes mutations liées à l’augmentation du trafic età l’évolution des outils permettant d’assurer le service. L’augmentation du nombre d’avions gérés parle contrôleur implique un nombre beaucoup plus important qu’auparavant d’informations à traiter etmémoriser ; or une part importante des causes d’incidents est déjà due à des problèmes de perception et demémorisation des informations amplifiées par la taille des écrans de contrôle. Ceux-ci sont particulièrementcritiques dans le cas de la perception des alarmes et des avertissements donnés par le système surla visualisation radar. Le design actuel de ce type d’alertes en France qui n’utilise que la couleur, faitaujourd’hui débat et il a été recommandé dans un bulletin de sécurité aérienne de le faire évoluer. Laperception des informations nécessaires à l’établissement d’une bonne conscience de la situation aériennedes contrôleurs est au centre de cette thèse. L’objet principal de ce travail est d’améliorer la conscience dela situation des contrôleurs en s’assurant que les éléments pertinents à leur disposition dont les alarmessont perçus dans des délais conformes à leur importance et que les actions nécessaires qui en découlentsont bien réalisées. Pour répondre à cette problématique, notre travail s’est porté sur deux axes. Le premierconsiste à étudier la tâche des contrôleurs de façon à analyser les informations nécessaires à la réalisationde certains objectifs et proposer un agent de suivi de la tâche qui pourrait les épauler. Le second porte surla perception des notifications, plus particulièrement en vision périphérique, et la conception de designspropres à améliorer leur perception ainsi que l’étude de leur impact sur la réalisation de la tâche. Grâceà une approche holistique basée sur l’utilisation d’un micro-monde ATC (Laby) et l’utilisation de capteursphysiologiques, nous avons pu évaluer plusieurs designs de notifications. Deux expérimentations ont étéconduites, la première utilisant de l’oculométrie, visait à mesurer le pouvoir attentionnel de cinq designs etleur impact sur la réalisation de la tâche. La seconde, en utilisant des données neurophysiologiques, visaità mesurer l’impact de deux designs sur la charge de travail. Les designs évalués ont servi à concevoir unsystème de notification intégré dans une nouvelle supervision radar qui a été couplé à l’agent de suivi dela tâche. La dernière expérimentation conduite durant ces travaux avait pour objectif d’évaluer l’impact decette nouvelle position de contrôle dont le design est orienté conscience de la situation sur la détection deproblèmes critiques. Les résultats obtenus montrent l’intérêt de ces nouveaux outils et leur impact positifsur la réalisation de la tâche des contrôleurs ainsi que la nécessité d’évaluer dans un contexte expérimentalcontrôlé les caractéristiques des designs de notifications pour les visualisations de supervision. / Air traffic control is undergoing a great change due to the increase of traffic and the evolution of thecontrol tools. The greater number of aircrafts managed by the controller implies a much greater load ofinformation to deal with and memorize than before. A significant part of accidents’ causes is already dueto the problem of information perception and memorization which is worsened by the size of the controlscreens which are particularly critical concerning the perception of alarms and warnings displayed onthe radar image. The current design in France of those types of alerts which rely on color is controversialand an Aviation Safety bulletin recommended it should be improved. The perception of informationcontributing to a satisfactory situation awareness by the air traffic controllers is central to this thesis whichaims at making sure alarms and relevant information are detected early enough and according to prioritiesand that the corresponding actions are actually performed. To answer the issue at stake our first work focusdealt with the study of the controllers’ task so as to analyze the required information in order to achievecertain objectives and provide a task monitoring agent that could support them. The second work focusdealt with the notifications perception, more specifically in the peripheral vision and the conception ofdesigns such as to improve their perception and the study of their impact on the performance of the task.Thanks to a holistic approach based on the use of an ATC microworld (Laby) and the use of physiologicalsensors we managed to assess several notification designs. Several experiments have been conducted,one using eye-tracking aimed at measuring attention capacity of five designs and their impact on theperformance of the task. The second used neurophysiologic data and aimed at measuring the impact oftwo designs on the work load. The assessed designs enabled to conceive a notification system integratedin a new radar supervision together with the task monitoring agent. The last experiment conducted in thecourse of this work aimed at measuring the impact of this new position of control whose design is situationawareness oriented, on the detection of critical problems. The obtained results show the relevance ofthese new tools and their positive impact on the performance of the task by controllers as well as the needto assess, in a controlled experimental context, the characteristics of notification designs for supervisionvisualization.