Dans le domaine de la vision par ordinateur l'utilisation de points d'intérêt (PI) est récurrente pour les problématiques de reconnaissance et de suivi d'objets. Plusieurs études ont prouvé l'utilité de ces techniques, associant robustesse et un temps de calcul pouvant être compatible avec le temps réel. Cette thèse propose d'étudier et d'exploiter ces descripteurs de statistiques de l'image sous un tout autre regard. Ainsi, nous avons menée une étude sur le lien entre les PI et la saillance visuelle humaine. De cette étude nous avons pu développer une méthode de prédiction de carte de saillance exploitant la rapidité d'exécution de ces détecteurs. Nous avons également exploité le pouvoir descriptif de ces PI afin de développer de nouvelles métriques de qualité d'images. Grâce à des résultats encourageant en terme de prédiction de qualité perçue et la faible quantité d'information utilisée, nous avons pu intégrer notre métrique "QIP" dans une chaîne de transmission d'images sur réseau sans fil de type MIMO. L'ajout de cette métrique permet d'augmenter la qualité d'expérience en garantissant la meilleure qualité malgré les erreurs introduites par la transmission sans fil. Nous avons étendu cette étude, par l'analyse fine des statistiques structurelles de l'image et des migrations d'attributs afin de proposer un modèle générique de prédiction des dégradations. Enfin, nous avons été amenés à conduire diverses expériences psychovisuelles, pour valider les approches proposées ou dans le cadre de la normalisation de nouveaux standards du comité JPEG. Ce qui a mené à développer une application web dédiée à l'utilisation et la comparaison des métriques de qualité d'images. / In the field of computer vision, the use of interest points (IP) is very frequent for objects tracking and recognition. Several studies have demonstrated the usefulness of these techniques, combining robustness and complexity that can be compatible with the real time. This thesis proposes to explore and exploit these image statistical descriptors under a different angle. Thus, we conducted a study on the relationship between IP and human visual saliency. In this study, we developed a method for predicting saliency maps relying on the efficiency of the descriptors. We also used the descriptive power of the PI to develop new metrics for image quality. With encouraging results in terms of prediction of perceived quality and the reduced amount of used information, we were able to integrate our metric "QIP" in an image transmission framework over a MIMO wireless network. The inclusion of this metric can improve the quality of experience by ensuring the best visual quality despite the errors introduced by the wireless transmission. We have extended this study by deeply analyzing structural statistics of the image and migration attributes to provide a generic model for predicting impairments. Finally, we conducted various psychovisual experiments to validate the proposed approaches or to contribute to JPEG standard committee. This led to develop a web application dedicated to the benchmark of image quality metrics.

The software is a key differentiator and driver of innovation in the automotive industry. The major challenges for software development are increasing in complexity, shorter time-to-market, increase in development cost and demand of quality assurance. The complexity is increasing due to emission legislations, variants of product and new communication technologies being interfaced with the vehicle. The shorter development time is due to competition in the market, which requires faster feedback loops of verification and validation of developed functionalities. The increase in development cost is contributed by two factors; the first is pre-launch cost, this involves the cost of error correction in development stages. Another is post-launch cost; this involves warranty and guarantees cost. As the development time passes the cost of error correction also increases. Hence it is important to detect the error as early as possible. All these factors affect the software quality; there are several cases where Original Equipment Manufacturer (OEM) have callbacks their product because of the quality defect. Hence, there is increased in the requirement of software quality assurance. The solution for these software challenges can be the early quality evaluation in continuous integration framework environment. The most prominent in today\'s automotive industry AUTomotive Open System ARchitecture (AUTOSAR) reference architecture is used to describe software component and interfaces. AUTOSAR provides the standardised software component architecture elements. It was created to address the issues of growing complexity; the existing AUTOSAR environment does have software quality measures, such as schema validations and protocols for acceptance tests. However, it lacks the quality specification for non-functional qualities such as maintainability, modularity, etc. The tool is required which will evaluate the AUTOSAR based software architecture and give the objective feedback regarding quality. This thesis aims to provide the quality measurement tool, which will be used for evaluation of AUTOSAR based software architecture. The tool reads the AUTOSAR architecture information from AUTOSAR Extensible Markup Language (ARXML) file. The tool provides configuration ability, continuous evaluation and objective feedback regarding software quality characteristics. The tool was utilised on transmission control project, and results are validated by industry experts.

The use of Virtual Reality (VR) is growing in commercial use, one type of VR headset, called mobile stand-alone system has limited resources for computing and memory. Because of this, when developing real-time applications for this type of VR headset, performance needs to be heavily considered. One popular optimization technique is Level-of-Detail (LOD) which is a technique that represents a model at various resolutions. One type of LOD is called continuous LOD which can represent a model at a continuous spectrum of detail, this is not frequently used, however, because of being less intuitive and more difficult to implement than other versions. This project researched a type of continuous LOD used with a new metric called Pixel Quality Per Meter (PQPM). PQPM relies on having a minimum edge length calculated using a model's screen coverage in relation to its size. To answer whether PQPM can be used together with continuous LOD for intuitive, simple, and efficient updating and rendering in VR. The continuous LOD only uses one low-poly mesh which is tessellated with the help of a Vertex Displacement Map (VDM) to the desired quality. This approach is then evaluated using Nvidia FLIP, an image comparison application that emulates the human visual system. The result was an intuitive and easily implementable LOD which is used together with PQPM to decide the optimal quality given the models size and coverage on the screen. The usage of PQPM did not result in optimal quality at all distances, due to smaller segments being present, which could disappear completely at far distances. The continuous LOD combined with the PQPM did also not scale well but worked well at lower qualities. The study showed groundwork for how PQPM could work together with continuous LOD, it provides a more intuitive and easily implementable continuous LOD than previous approaches, however, because of the scalability issues, further work needs to be done to optimize this approach. / Användingen av Virtual Reality (VR) i kommersiellt syfte ökar, en typ av dessa VR headsets kallas mobila fristående system. Mobila fristående system har begränsade resurser när det kommer till att göra beräkningar och minne. På grund av detta måste man noga överväga prestanda. En populär optimeringsmetod kallas Level-of-Detail (LOD) vilket är en metod som kan representera en modell på ett antal olika upplösningar. En typ av LOD kallas kontinuerlig LOD som kan representera en modell med en kontinuerling detaljnivå. Denna används inte frekvent däremot på grund av att den är mindre intuitiv och svårare att implementera och uppdatera detaljnivå. Detta projektet undersökte en typ av kontinuerlig LOD där den används tillsammans med ett nytt metriskt mått som kallas Pixel Quality Per Meter (PQPM). PQPM byggs på att beräkna en minimal kantlängd via hur stor del av skärmen den täcker i relation till dess storlek. Detta undersöks för att besvara ifall PQPM kan användas tillsammans med kontinuerlig LOD för intuitiv, simpel och effektiv uppdatering och rendering i VR. Den kontinuerliga LOD använder sig bara av en low-poly mesh som är tessellerad med hjälp av en Vertex Displacement Map (VDM) till en önskad kvalitet. Detta evalueras sedan genom Nvidia FLIP, en bildjämförelseapplikation som emulerar människosyn. Resultatet var en intuitiv och enkelt implementerbar kontinuerlig LOD som användes tillsammans med PQPM för att bestämma den optimala detaljnivån baserat på modellens storlek och hur mycket av skärmen den täckte. Användingen av PQPM resulterade inte i optimal kvalitet för alla distanser, på grund av mindre segment som kunde försvinna helt på längre distanser. Den kontinuerliga LOD tillsammans med PQPM hade dålig skalning i prestanda men fungerade bra vid lägre kvalitetsnivåer. Studien visa grundarbetet för hur PQPM kan användas tillsammans med kontinuerlig LOD. Detta tillvägagångsättet tillät för ett mer intuitivt och simpelt sätt att implementera än tidigare tillvägagångssätt. Däremot, på grund av skalningsproblemen, fortsatt arbete måste göras för att optimera detta tillvägagångssättet.

The use of audio and video communication has increased exponentially over the last decade and has gone from speech over GSM to HD resolution video conference between continents on mobile devices. As the use becomes more widespread the interest in delivering high quality media increases even on devices with limited resources. This includes both development and enhancement of the communication chain but also the topic of objective measurements of the perceived quality. The focus of this thesis work has been to perform enhancement within speech encoding and video decoding, to measure influence factors of audio and video performance, and to build methods to predict the perceived video quality. The audio enhancement part of this thesis addresses the well known problem in the GSM system with an interfering signal generated by the switching nature of TDMA cellular telephony. Two different solutions are given to suppress such interference internally in the mobile handset. The first method involves the use of subtractive noise cancellation employing correlators, the second uses a structure of IIR notch filters. Both solutions use control algorithms based on the state of the communication between the mobile handset and the base station. The video enhancement part presents two post-filters. These two filters are designed to improve visual quality of highly compressed video streams from standard, block-based video codecs by combating both blocking and ringing artifacts. The second post-filter also performs sharpening. The third part addresses the problem of measuring audio and video delay as well as skewness between these, also known as synchronization. This method is a black box technique which enables it to be applied on any audiovisual application, proprietary as well as open standards, and can be run on any platform and over any network connectivity. The last part addresses no-reference (NR) bitstream video quality prediction using features extracted from the coded video stream. Several methods have been used and evaluated: Multiple Linear Regression (MLR), Artificial Neural Network (ANN), and Least Square Support Vector Machines (LS-SVM), showing high correlation with both MOS and objective video assessment methods as PSNR and PEVQ. The impact from temporal, spatial and quantization variations on perceptual video quality has also been addressed, together with the trade off between these, and for this purpose a set of locally conducted subjective experiments were performed.

Cette thèse porte sur l’étude des stratégies d’amélioration de la qualité d’image dans les systèmes de communication sans fil et sur la conception de nouvelles métriques d’évaluation de la qualité. Tout d'abord, une nouvelle métrique de qualité d'image à référence réduite, basée sur un modèle statistique dans le domaine des ondelettes complexes, a été proposée. Les informations d’amplitude et de phase relatives des coefficients issues de la transformée en ondelettes complexes sont modélisées à l'aide de fonctions de densité de probabilité. Les paramètres associés à ces fonctions constituent la référence réduite qui sera transmise au récepteur. Ensuite, une approche basée sur les réseaux de neurones à régression généralisée est exploitée pour construire la relation de cartographie entre les caractéristiques de la référence réduite et le score objectif.Deuxièmement, avec la nouvelle métrique, une nouvelle stratégie de décodage est proposée pour la transmission d’image sur un canal de transmission sans fil réaliste. Ainsi, la qualité d’expérience (QoE) est améliorée tout en garantissant une bonne qualité de service (QoS). Pour cela, une nouvelle base d’images a été construite et des tests d’évaluation subjective de la qualité de ces images ont été effectués pour collecter les préférences visuelles des personnes lorsqu’elles sélectionnent les images avec différentes configurations de décodage. Un classificateur basé sur les algorithmes SVM et des k plus proches voisins sont utilisés pour la sélection automatique de la meilleure configuration de décodage.Enfin, une amélioration de la métrique a été proposée permettant de mieux prendre en compte les spécificités de la distorsion et la préférence des utilisateurs. Pour cela, nous avons combiné les caractéristiques globales et locales de l’image conduisant ainsi à une amélioration de la stratégie de décodage.Les résultats expérimentaux valident l'efficacité des métriques de qualité d'image et des stratégies de transmission d’images proposées. / This thesis focuses on the study of image quality strategies in wireless communication systems and the design of new quality evaluation metrics:Firstly, a new reduced-reference image quality metric, based on statistical model in complex wavelet domain, has been proposed. The magnitude and the relative phase information of the Dual-tree Complex Wavelet Transform coefficients are modelled by using probability density function and the parameters served as reduced-reference features which will be transmitted to the receiver. Then, a Generalized Regression Neural Network approach is exploited to construct the mapping relation between reduced-reference feature and the objective score.Secondly, with the new metric, a new decoding strategy is proposed for a realistic wireless transmission system, which can improve the quality of experience (QoE) while ensuring the quality of service (QoS). For this, a new database including large physiological vision tests has been constructed to collect the visual preference of people when they are selecting the images with different decoding configurations, and a classifier based on support vector machine or K-nearest neighboring is utilized to automatically select the decoding configuration.Finally, according to specific property of the distortion and people's preference, an improved metric has been proposed. It is the combination of global feature and local feature and has been demonstrated having a good performance in optimization of the decoding strategy.The experimental results validate the effectiveness of the proposed image quality metrics and the quality strategies.