No description available.

This work optimizes a CMOS image pixel sensor circuit for being used in a compressive sensing (CS) image sensor. The CS image sensor sums neighbor pixel outputs and hence reduces analog to digital conversions. Efforts are also made to improve the circuit that performs such pixel summation. With the optimized design, a CMOS image sensor circuit with a compression ratio of 4 is designed using a 130 nm CMOS technology from Global foundries. The design pixel sensor has a 256X256 pixel array. Simulation shows that the developed image sensors can achieve peak signal to noise ratio (PSNR) of 28 dB and 37.8 dB for benchmark images Cameraman and Lenna, respectively.

. / The development of high-resolution image sensors with smaller pixel sizes is facing critical issues, such as optical and electrical crosstalk, dark current and dynamic range. As part of this thesis, we addressed this issue by proposing the integration of MOS capacitor deep trench isolation (CDTI). Our studies focus on the validation of the proposal with the aim of improving performances compared to the state of the art. First, we modeled interface states Si/SiO2 and the charge in the oxide. By TCAD simulations, using our model, we were able to evaluate the main characteristics of a pixel. We have validated this approach by comparison between simulations and measurements on a 1.4μm DTI pixel. Then, we developed manufacturing processes for integrating CDTI and defined the associated key parameters. With TCAD simulations of process type, we could achieve the desired performances while keeping a short development cycle and cost. Finally, we have designed, manufactured and tested a 1.4μm CDTI pixel ; we got a very low dark current: ~ 1 aA/pixel at 60°C, which is 6 times less than the DTI pixel, and doubled saturation charge up to 12000e-. Other performances are comparable between the two types of pixels. We have demonstrated the validity of the proposed CDTI solution CDTI

No description available.

Les réseaux de capteurs sans fil sont conçus dans le but de remplir différentes tâches de surveillance dans des conditions environnementales variées. Ces petits appareils électroniques sont capables de détecter, traiter et transmettre des données environnementales avec des communications multi-sauts et peuvent par conséquent aussi se coordonner. En même temps, ces dispositifs ont des ressources limitées (mémoire, capacités de calcul) et doivent fonctionner le plus souvent sur des batteries. C’est pour ces raisons que les recherches menées dans le domaine des réseaux de capteurs possèdent naturellement une forte partie qui concerne la réduction de la consommation d’énergie et une auto-organisation du réseau.Nos recherches considèrent les applications critiques de surveillance. Ces applications peuvent avoir des exigences très différentes des réseaux de capteurs traditionnels. De plus, nous utilisons des capteurs images, dont l’activité est définie en fonction de la criticité de l’application. Un ordonnancement basé sur la criticité permet de définir des nœuds sentinelles qui possèderont une vitesse de capture plus grande, cela afin d’avoir une probabilité plus élevée de détecter des intrusions et d’alerter leurs nœuds voisins. Au niveau de la couche de contrôle d’accès au medium (couche MAC), les approches alternant activité-sommeil (consistant à allumer et éteindre la radio de manière cyclique) sont utilisées pour préserver l’énergie et prolonger la durée de vie du réseau. Cependant, tout en conservant l’énergie, les applications critiques de surveillance ne doivent pas compromettre la qualité de la surveillance et le réseau doit être toujours en mesure de propager rapidement les messages d’alerte. Notre but est de définir un protocole MAC qui pourrait réduire la latence de propagation d’alerte ainsi que de prolonger la durée de vie du réseau. Nous proposons tout d’abord une approche originale pour déterminer dynamiquement la durée de la période d’activité radio des nœuds pour augmenter la probabilité de propager rapidement les alertes. Les résultats des simulations ont confirmé que notre approche réussit à améliorer la réactivité du système par rapport à une approche statique. En même temps, notre proposition permet de réduire considérablement la consommation d’énergie du réseau. Ensuite, nous avons implémenter notre approche sur des capteurs réels et les résultats obtenus sont très proches des résultats de simulation. / Wireless Sensor Networks (WSNs) are designed for the purpose of completing different monitoring tasks under various environmental conditions. The small electronic devices called sensors are capable of sensing, processing and communicating the environmental data through multi-hop communication and coordination. These devices have limited resources (memory, computing capabilities) and usually run on batteries. This is the reason the research on wireless sensor networks have been focused on energy efficiency and self-organization of the network. We consider mission-critical surveillance applications in our research work. These applications can have different requirements than traditional WSNs. In addition, we use image sensor nodes, whose activity is defined based on criticality of the application. The criticality-based scheduling scheme defines sentry nodes with faster capture rates, to have higher probability to detect intrusions and to alert neighbor nodes. At Medium Access Control (MAC) Layer level, duty cycled approaches are used to preserve energy and prolong the network lifetime. However, while conserving energy, mission-critical surveillance applications cannot compromise on quality of surveillance and the network should still be able to quickly propagate the alert messages. In this thesis, we propose a low latency, energy efficient adaptive MAC protocol. We first propose an original approach to dynamically determine the duty-cycle length of sensor nodes to increase the probability of quick propagation of alerts. Simulation results confirmed that our approach succeeds in improving the system responsiveness when compared to a static duty-cycling approach. At the same time, our proposition considerably reduces the energy consumption of the network. Then, we implemented our approach on sensor node hardware and results were found to be very close to our simulation results.

La taille des pixels des capteurs d’image CMOS approche aujourd’hui lemicron. Dans ce contexte, le courant d’obscurité reste un paramètrecritique. Il se superpose au courant photogénéré en affectant la qualité del’image par l’apparition de pixels blancs. La contamination métalliqueintroduite au cours du procédé de fabrication joue un rôle prépondérant dansla création des défauts à l’origine de ce courant d’obscurité. Cette étude apermis d’établir les seuils de dangerosité de différents élémentsmétalliques sur la technologie imageur. L’origine de contaminationsaccidentelles a été identifiée lors de crises de rendement. Pour cela, untravail sur les techniques de détection a été mené par µPCD, DLTS, pompagede charge, SIMS, TEM et photoluminescence. La spectroscopie de courantd’obscurité (DCS), particulièrement efficace dans ce contexte, a étédéveloppée pour l’identification de contaminations en or, tungstène etmolybdène, avec des limites de détection qui atteignent 108 à 1010 at/cm3.Nous observons la quantification du courant d’obscurité et étudionsl’amplification du champ électrique sur le taux de génération afin demodéliser les pics de courant d’obscurité obtenus. Le comportement decertains métaux dans le silicium est précisé par ces expériences, et nousévaluons l’efficacité de piégeage de plusieurs substrats imageur. Ce travailconduit à la mise en place de protocoles de contrôle de la contaminationmétallique en salle blanche. / Pixels size of CMOS image sensors is now decreasing towards one micron. Inthat context, dark current is a critical parameter. It superimposes with thecurrent generated by photons and affects the image quality with whitepixels. The metallic contamination introduced during the fabrication processplays an important role in the generation of defects that induce this darkcurrent. This study has allowed to determine dangerousness thresholds ofseveral metals on the imager technology. The origin of some accidentalcontaminations has been identified during yield crisis. Some work withdetection techniques has been performed with µPCD, DLTS, charge pumping,SIMS, TEM and photoluminescence. Dark current spectroscopy (DCS),particularly adapted to this situation, has been developped for theidentification of gold, tunsgten and molybdenum contaminations, withdetection limits that reach 108 to 1010 at/cm3. We have observed the darkcurrent quantization and studied the electric field enhancement ofgeneration rate to model the dark current peaks obtained. The behavior ofsome metals in silicon is confirmed by these experiments and we haveevaluated the getter efficiency of different substrates for image sensors.This work has lead to the application of protocols for the metalliccontamination control in clean room.

The image sensors embedded in the instruments on board of different satellites are a key aspect in the success of a mission of any type such as Earth observation, astronomy or planetary and universe exploration. The quality of the sensor itself as well as the electronics embedded in the instrument suffer from aging, not only due to the usage, but also due to the space environment factors. The outer space harsh environment is composed of charged particles with an incredibly high energy that can get through the shielding of the spacecraft thus causing damage to the sensors, or any part of the spacecraft. After a satellite is placed in orbit, there is no other way to check on its condition than to analyse the data transmitted to the ground stations.   It is well known that we can measure the evolution of the instrument’s detection unit by measuring the variation of the quality of the images taken in orbit. As it has been done in past work, the atypical behaviour can be detected at a pixel level by analysing the images taken in orbit, preferably in complete darkness to discard any parasitic light. Past analysis of the dark images from the sensors on board the Picard, SPOT4 and Sentinel2A mission, proposed a classification of the different atypical behaviour that a pixel can show, such as hot, transitory-hot or RTS. By the end of the mission this defects may even be present in several pixels.   The purpose of this work is to process the dark images obtained on orbit, at the end of the mission PLEIADES, to better understand the behaviour both Panchromatic and Multispectral CCD sensors and the differences in performance with respect to the tests made previously on ground. For this, a description of the framework of the PLEIADES mission and the technical design aspects of the CCD sensors on board are presented. Subsequently, the different image processing techniques used and developed are described, along with the tests made to get to the positive conclusion of no major anomalies detected in the PLEIADES images sensors. The analysis is done by means of new set of functions adjusted to the structure of the data, as well as the same detection software used in the previous analysis. Finally, a discussion is presented along with a further perspective for the analysis of future sets of data.

<p>Imaging in low light is difficult because the number of photons arriving at the image sensor is low. This is a major technological challenge for applications such as surveillance and autonomous driving. Conventional CMOS image sensors (CIS) circumvent this issue by using techniques such as burst photography. However, this process is slow and it does not solve the underlying problem that the CIS cannot efficiently capture the signals arriving at the sensors. This dissertation focuses on solving this problem using a combination of better image sensors (Quanta Image Sensors) and computational imaging techniques.</p> <p><br></p> <p>The first part of the thesis involves understanding how the quanta image sensors work and how they can be used to solve the low light imaging problem. The second part is about the algorithms that can deal with images obtained in low light. The contributions in this part include – 1. Understanding and proposing solutions for the Poisson noise model, 2. Proposing a new machine learning scheme called student-teacher learning for helping neural networks deal with noise, and 3. Developing solutions that work not only for low light but also for a wide range of signal and noise levels. Using the ideas, we can solve a variety of applications in low light, such as color imaging, dynamic scene reconstruction, deblurring, and object detection.</p>

A retinografia e angiografia digital são técnicas de observação e captura de imagens do fundo do olho que permitem o diagnóstico de diversas patologias. O aperfeiçoamento dos retinógrafos digitais, viabilizado pelos avanços tecnológicos observados nas últimas décadas, tornou possíveis modificações sistêmicas inovadoras como a apresentada e discutida neste trabalho. O principal objetivo é avaliar comparativamente a resolução espacial de imagens captadas com uma configuração que utiliza apenas um sensor monocromático e um conjunto de LEDs cromáticos, com imagens geradas por um sensor com filtro cromático integrado. A proposta é suportada por resultados teóricos que indicam um desempenho superior do uso de sensores monocromáticos em retinografia. Para validação da nova metodologia o sistema descrito foi montado e várias fotos da retina foram capturadas. Todo o processo de imageamento do fundo do olho é descrito, incluindo rotinas de software que precisaram ser criadas em decorrência das inovações. A partir das imagens capturadas foram aplicadas técnicas de avaliação global de maneira que as previsões teóricas puderam ser verificadas. Por fim são apresentadas conclusões sobre o desempenho global do sistema e adicionados tópicos como sugestões para continuidade futura do trabalho. / The digital retinography and angiography are techniques to observe and capture images of the eye fundus that allows the diagnosis of several diseases. The improvements of digital fundus camera made possible by technological advances seen in recent decades have made possible innovative systemic changes, as the presented and discussed in this work. The main objective is to comparatively evaluate the spatial resolution of images captured with a configuration that uses only a monochrome sensor and a set of chromatic LEDs, with images generated by a sensor with integrated color filter. The proposal is supported by theoretical results that indicate a superior performance when using a monochrome sensor to get fundus images. To validate the new methodology, the system described was assembled and several retina pictures were taken. The whole process of imaging the eye fundus is described, including software procedures that had to be created as a consequence of the innovations. From the images captured, overall evaluation techniques have been applied, so that theoretical forecasts could be verified. Finally, conclusions are presented on the overall performance of the system and suggestions for topics for future work were added.

A retinografia e angiografia digital são técnicas de observação e captura de imagens do fundo do olho que permitem o diagnóstico de diversas patologias. O aperfeiçoamento dos retinógrafos digitais, viabilizado pelos avanços tecnológicos observados nas últimas décadas, tornou possíveis modificações sistêmicas inovadoras como a apresentada e discutida neste trabalho. O principal objetivo é avaliar comparativamente a resolução espacial de imagens captadas com uma configuração que utiliza apenas um sensor monocromático e um conjunto de LEDs cromáticos, com imagens geradas por um sensor com filtro cromático integrado. A proposta é suportada por resultados teóricos que indicam um desempenho superior do uso de sensores monocromáticos em retinografia. Para validação da nova metodologia o sistema descrito foi montado e várias fotos da retina foram capturadas. Todo o processo de imageamento do fundo do olho é descrito, incluindo rotinas de software que precisaram ser criadas em decorrência das inovações. A partir das imagens capturadas foram aplicadas técnicas de avaliação global de maneira que as previsões teóricas puderam ser verificadas. Por fim são apresentadas conclusões sobre o desempenho global do sistema e adicionados tópicos como sugestões para continuidade futura do trabalho. / The digital retinography and angiography are techniques to observe and capture images of the eye fundus that allows the diagnosis of several diseases. The improvements of digital fundus camera made possible by technological advances seen in recent decades have made possible innovative systemic changes, as the presented and discussed in this work. The main objective is to comparatively evaluate the spatial resolution of images captured with a configuration that uses only a monochrome sensor and a set of chromatic LEDs, with images generated by a sensor with integrated color filter. The proposal is supported by theoretical results that indicate a superior performance when using a monochrome sensor to get fundus images. To validate the new methodology, the system described was assembled and several retina pictures were taken. The whole process of imaging the eye fundus is described, including software procedures that had to be created as a consequence of the innovations. From the images captured, overall evaluation techniques have been applied, so that theoretical forecasts could be verified. Finally, conclusions are presented on the overall performance of the system and suggestions for topics for future work were added.