0000000000146598
AUTHOR
Dominique Ginhac
Smart cameras on a chip: using complementary metal-oxide-semiconductor (CMOS) image sensors to create smart vision chips
Abstract: In this chapter, we introduce the fundamental concept of smart cameras on a chip or smart vision chips that simultaneously integrate the same die image capture capability and highly complex image processing. Successive technology scaling has made possible the integration of specific processing elements designed at chip level, at column level or at pixel level. To illustrate this continuous evolution, we survey three different categories of vision chips, exploring first the pioneering works on artificial retinas, then describing the most significant computational chips, and finally presenting the most recent image processing chips able to perform complex algorithms at a high frame …
An FPGA-based design for real-time Super Resolution Reconstruction
Since several decades, the camera spatial resolution is gradually increasing with the CMOS technology evolution. The image sensors provide more and more pixels, generating new constraints for the suitable optics. As an alternative, promising solutions propose Super Resolution (SR) image reconstruction to extend the image size without modifying the sensor architecture. Convincing state-of art studies demonstrate that these methods could even be implemented in real-time. Nevertheless, artifacts can be observed in highly textured areas of the image. In this paper, we propose a Local Adaptive Spatial Super Resolution (LASSR) method to fix this limitation. A real-time texture analysis is include…
Implicit learning of a repeated segment in continuous tracking: A reappraisal
Several prior studies (e.g., Shea, Wulf, Whitacre, & Park, 2001; Wulf & Schmidt, 1997) have apparently demonstrated implicit learning of a repeated segment in continuous-tracking tasks. In two conceptual replications of these studies, we failed to reproduce the original findings. However, these findings were reproduced in a third experiment, in which we used the same repeated segment as that used in the Wulf et al. studies. Analyses of the velocity and the acceleration of the target suggests that this repeated segment could be easier to track than the random segments serving as control, accounting for the results of Wulf and collaborators. Overall these experiments suggest that lea…
Efficient smart-camera accelerator: A configurable motion estimator dedicated to video codec
Smart cameras are used in a large range of applications. Usually the smart cameras transmit the video or/and extracted information from the video scene, frequently on compressed format to fit with the application requirements. An efficient hardware accelerator that can be adapted and provide the required coding performances according to the events detected in the video, the available network bandwidth or user requirements, is therefore a key element for smart camera solutions. We propose in this paper to focus on a key part of the compression system: motion estimation. We have developed a flexible hardware implementation of the motion estimator based on FPGA component, fully compatible with…
Action, observation et imagerie mentale : apports de l'apprentissage implicite au domaine moteur
International audience
Procedural variations around a SRT task
International audience
RGB-Event Fusion for Moving Object Detection in Autonomous Driving
Moving Object Detection (MOD) is a critical vision task for successfully achieving safe autonomous driving. Despite plausible results of deep learning methods, most existing approaches are only frame-based and may fail to reach reasonable performance when dealing with dynamic traffic participants. Recent advances in sensor technologies, especially the Event camera, can naturally complement the conventional camera approach to better model moving objects. However, event-based works often adopt a pre-defined time window for event representation, and simply integrate it to estimate image intensities from events, neglecting much of the rich temporal information from the available asynchronous ev…
Efficient smart-camera accelerator: an configurable motion estimator dedicated to video codec
International audience; Smart cameras are used in a large range of applications. Usually the smart cameras transmit the video or/and extracted information from the video scene, frequently on compressed format to fit with the application requirements. An efficient hardware accelerator that can be adapted and provide the required coding performances according to the events detected in the video, the available network bandwidth or user requirements, is therefore a key element for smart camera solutions. We propose in this paper to focus on a key part of the compression system: motion estimation. We have developed a flexible hardware implementation of the motion estimator based on FPGA componen…
Modular VLIW processor based on FPGA for real-time image processing
National audience
Real-time High Dynamic Range based on Multiple Non Destructive ReadOut during a Single Exposure
This paper presents a new method based on Non Destructive Readout (NDRO) to improve multi-exposure High Dynamic Range (HDR) Imaging. A sequence of Low-Dynamic Range (LDR) images can then be acquired during a single exposure. The concept enables the latency between LDR images to be removed as well as the intrinsic ghost artifacts observed using state-of-art HDR systems based on multi-exposures. The method has been applied to improve the performances of HDR sensor based on logarithmic pixels. Using the NDRO method, a Short Wave InfraRed (SWIR) camera has been designed to produce HDR IR videos. A real-time HDR video stream generation is achieved based on GPU implantation.
A 1.3 megapixel FPGA-based smart camera for high dynamic range real time video
International audience; A camera is able to capture only a part of a high dynamic range scene information. The same scene can be fully perceived by the human visual system. This is true especially for real scenes where the difference in light intensity between the dark areas and bright areas is high. The imaging technique which can overcome this problem is called HDR (High Dynamic Range). It produces images from a set of multiple LDR images (Low Dynamic Range), captured with different exposure times. This technique appears as one of the most appropriate and a cheap solution to enhance the dynamic range of captured environments. We developed an FPGA-based smart camera that produces a HDR liv…
Machine Learning VS Transfer Learning - Smart Camera Implementation for Face Authentication
The aim of this paper is to highlight differences between classical machine learning and transfer learning applied to low cost real-time face authentication. Furthermore, in an access control context, the size of biometric data should be minimized so it can be stored on a remote personal media. These constraints have led us to compare only lightest versions of these algorithms. Transfer learning applied on Mobilenet v1 raises to 85% of accuracy, for a 457Ko model, with 3680s and 1.43s for training and prediction tasks. In comparison, the fastest integrated method (Random Forest) shows accuracy up to 90% for a 7,9Ko model, with a fifth of a second to be trained and a hundred of microseconds …
Exploiting Semantic Trajectories Using HMMs and BIM for Worker Safety in Dynamic Environments
International audience; Understanding dynamic behaviors of moving objects using positioning technologies for construction safety monitoring is still an open research issue. One task; that is a small subset in the widespread field of objects dynamics is the enrichment of the location data of users with the semantic information for studying their mobility patterns in the context of the environment. However, incorporating the semantics related to the environment gets complex in case of the dynamic construction sites where the site spaces are kept evolving with time. For instance, new walls and infrastructure supports are added often on sites, while others are detached. Similar situations open …
A Review on Applications of Big Data for Disaster Management
International audience; The term " disaster management " comprises both natural and man-made disasters. Highly pervaded with various types of sensors, our environment generates large amounts of data. Thus, big data applications in the field of disaster management should adopt a modular view, going from a component to nation scale. Current research trends mainly aim at integrating component, building, neighborhood and city levels, neglecting the region level for managing disasters. Current research on big data mainly address smart buildings and smart grids, notably in the following areas: energy waste management, prediction and planning of power generation needs, improved comfort, usability …
Evaluation de l'outil SynDEx en vue de prototypage rapide d'applications de traitement d'images sur machine MIMD-DM
International audience; The goal of this paper is to evaluate the SynDEx system-level CAD tool in order to estimate its usefulness for fast prototyping of image processing applications on a MIMD-DM architecture. This software can assist the programmer during the implementation of image processing applications in his constrained search for an efficient matching between algorithm and architecture. Two main conclusions were drawn from this work. First, the implementation of a connected component labeling algorithm on a multi-transputer architecture allowed us to quantify the gap between the estimated performances predicted by SynDEx and the effective performances measured on the generated exec…
The Gamma Cube: a new way to explore the gamma-ray sky
International audience; We propose a new concept to allow the tracking of electrons in a gamma-ray telescope operating in the 5–100 MeV band. The idea of this experiment is to image the ionizing tracks that charged particles produce in a scintillator. It is a pair creation telescope at high energy and a Compton telescope with electron tracking at low energy. The telescope features a large scintillator transparent to the scintillation light, an ad-hoc optical system and a high resolution and highly sensitive imager. The performance perspectives and the advantages of such a system are outstanding but the technical difficulties are serious. A few years of research and development within the sc…
Emergence of chunks during a 1023-alternative reaction time task
International audience
A critical review on the implementation of static data sampling techniques to detect network attacks
International audience; Given that the Internet traffic speed and volume are growing at a rapid pace, monitoring the network in a real-time manner has introduced several issues in terms of computing and storage capabilities. Fast processing of traffic data and early warnings on the detected attacks are required while maintaining a single pass over the traffic measurements. To palliate these problems, one can reduce the amount of traffic to be processed by using a sampling technique and detect the attacks based on the sampled traffic. Different parameters have an impact on the efficiency of this process, mainly, the applied sampling policy and sampling ratio. In this paper, we investigate th…
An affordable contactless security system access for restricted area
International audience; We present in this paper a security system based on identity verification process and a low-cost smart camera , intended to avoid unauthorized access to restricted area. The Le2i laboratory has a longstanding experience in smart cameras implementation and design [1], for example in the case of real-time classical face detection [2] or human fall detection [3]. The principle of the system, fully thought and designed in our laboratory, is as follows: the allowed user presents a RFID card to the reader based on Odalid system [4]. The card ID, time and date of authorized access are checked using connection to an online server. In the same time, multi-modality identity ve…
High Dynamic Range Real-time Vision System for Robotic Applications
International audience; Robotics applications often requires vision systems capable of capturing a large amount of information related to a scene. With many camera sensors, the perception of information is limited in areas with strong contrasts. The High Dynamic Range (HDR) vision system can deal with these limitations. This paper describes the HDR-ARtiSt hardware platform (High Dynamic Range Advanced Real-time imaging System), a FPGA-based architecture that can produce a real- time high dynamic range video from successive image acquisition.
Understanding the human behaviors in dynamic urban areas of interest
International audience; The fast growth of urban populations across evolving cities is resulting in new types of technical, physical, material, and safety-related challenges and constraints (Chen et al. 2019). Humans perceive, move, and perform actions (called behaviors) in different urban areas, and make interactions with the objects, the other humans, and the environment for achieving their goals for attaining the desired level of satisfaction (Wagner et al., 2018; Chen et al., 2015). More formally, human behaviors are observable actions or reactions of humans in response to external or internal stimuli (Wagner et al., 2018; Hong et al., 2015; Yan et al., 2015). Over the last decade, huma…
Incidental learning of interactions between motor and linguistic sequences
International audience
HDR-ARtiSt: High Dynamic Range Advanced Real-Time Imaging System
International audience; This paper describes the HDR-ARtiSt hardware platform, a FPGA-based architecture that can produce a real- time high dynamic range video from successive image acquisition. The hardware platform is built around a standard low dynamic range (LDR) CMOS sensor and a Virtex 5 FPGA board. The CMOS sensor is a EV76C560 provided by e2v. This 1.3 Megapixel device offers novel pixel integration/readout modes and em- bedded image pre-processing capabilities including multiframe acquisition with various exposure times. Our approach consists of a hardware architecture with different algorithms: double exposure control during image capture, building of an HDR image by combining the…
Implicit motor learning in discrete and continuous tasks: Toward a possible explanation of discrepant results
Abstract
Movement Behavior Analysis of Workers using Spatio-temporal Trajectories for Safety Management
International audience
Influence of the response-stimulus interval on implicit sequence learning: constant vs. variable RSIs
International audience
Apprentissage moteur implicite : variations autour d’une tâche de TRS
International audience
Design of low noise and low power photo-receptor for CMOS vision sensor
International audience
The role of time and pace in sequence learning
International audience
HDR-ARtiSt : une caméra intelligente dédiée à la vidéo à grande dynamique en temps réel
International audience; Une caméra vidéo standard n'est capable de capturer qu'une partie de l'information contenue dans une scène de gamme dynamique élevée. La même scène peut être presque entièrement perçue par le système visuel humain. Cela est particulièrement vrai pour les scènes réelles où la différence d'intensité lumineuse entre les zones sombres et les zones claires est élevée. La technique d'imagerie qui peut surmonter ce problème est appelé HDRi (High Dynamic Range imaging). Elle produit des images à partir d'un ensemble de plusieurs images LDRs (Low Dynamic Range), capturées avec différents temps d'exposition. Cette technique apparaît comme l'une des solutions la plus appropriée…
Adéquation Algorithme architecture : Aspects logiciels, matériels et cognitifs
Les travaux présentés dans le cadre de cette Habilitation à Diriger des Recherches s'inscrivent principalement dans la problématique dite d'" Adéquation Algorithme Architecture ". Ils ont pour objectif commun la mise en œuvre de systèmes matériels et logiciels dédiés à la vision artificielle à fortes contraintes temporelles. Ils se focalisent sur différents aspects cruciaux tels que l'acquisition d'images par des capteurs dédiés, le développement et la programmation d'architectures optimisées de traitement des images et l'implantation d'algorithmes de traitement du signal et d'images en temps réel sur ces architectures.
Implicit Motor Learning in discrete vs. continuous tasks
International audience
SiPM cryogenic operation down to 77 K
International audience; Silicon PhotoMultiplier (SiPM) is composed of extremely sensitive photosensors based on the Geiger Mode Avalanche PhotoDiode (GM-APD), which operate as a digital pixel sensitive to single photons. SiPMs are being considered for applications in low temperature environments, such as noble-liquid detectors for dark matter searches or neutrino physics and GM-APD is promising technology for space Compton telescopes. While it is well known that the dark count rate, one of the main limitations of SiPM, is reduced at low temperature, a detailed study of the behavior of the device in cryogenic environment is necessary to assess its performances. In this paper, we present meas…
WiseEye: A Platform to Manage and Experiment on Smart Camera Networks
International audience; Embedded vision is probably at the edge of phenomenal expansion. The smart cameras are embedding some processing units which are more and more powerful. Last decade, high-speed image processing can be implemented on specifically designed architectures [1] nevertheless the designing time of such systems was quite high and time to market therefore as well. Since, powerful chips (i.e System On Chip) and quick prototyping methodologies are contently emerging [2],[3],[4] and enable more complex algorithms to be implemented faster. Moreover, smart cameras which are embedding flexible and powerful multi-core processors or Graphic Processors Unit (GPU) are now available and …
Multi-pixel Geiger mode imager for medical applications
International audience; Nowadays, there are two types of sensors to detect the low luminous flux, PMT (Photomultiplier Tube) and Geiger-APD (Geiger Avalanche Photodiode). The domain of Geiger-APD has reached an advanced development in the last years. The basic idea of this structure consists in polarizing an APD in Geiger-mode by applying a voltage beyond its breakdown voltage. In this case of polarization, the APD is working in a special mode and is able to detect the single photon. The theory of detection of single photon using this detector has been invented in the beginning of 90's and developed for detection of low light intensity. By using this kind of photodiode in the Geiger-mode, w…
Time Unification on Local Binary Patterns Three Orthogonal Planes for Facial Expression Recognition
International audience; Machine learning has known a tremendous growth within the last years, and lately, thanks to that, some computer vision algorithms started to access what is difficult or even impossible to perceive by the human eye. While deep learning based computer vision algorithms have made themselves more and more present in the recent years, more classical feature extraction methods, such as the ones based on Local Binary Patterns (LBP), still present a non negligible interest, especially when dealing with small datasets. Furthermore, this operator has proven to be quite useful for facial emotions and human gestures recognition in general. Micro-Expression (ME) classification is…
Comparative study of deep learning and classical methods applied to face authentication in context of high constraints application
International audience
Design And Characterization Of Automated Color Sensor System
Abstract The paper presents a color sensor system that can process light reflected from a surface and produce a digital output representing the color of the surface. The end-user interface circuit requires only a 3-bit pseudo flash analog-to-digital converter (ADC) in place of the conventional/typical design comprising ADC, digital signal processor and memory. For scalability and compactness, the ADC was designed such that only two comparators were required regardless of the number of color/wavelength to be identified. The complete system design has been implemented in hardware (bread board) and fully characterized. The ADC achieved less than 0.1 LSB for both INL and DNL. The experimental r…
Multiple representations and mechanisms for visuomotor adaptation in young children
International audience; In this study, we utilized transformed spatial mappings to perturb visuomotor integration in 5-yr-old children and adults. The participants were asked to perform pointing movements under five different conditions of visuomotor rotation (from 0° to 180°), which were designed to reveal explicit vs. implicit representations as well as the mechanisms underlying the visual-motor mapping. Several tests allowed us to separately evaluate sensorimotor (i.e., the dynamic dimension of movement) and cognitive (i.e., the explicit representations of target position and the strategies used by the participants) representations of visuo-proprioceptive distortion. Our results indicate…
Skeletons for parallel image processing: an overview of the SKiPPER project
International audience; This paper is a general overview of the SKIPPER project, run at Blaise Pascal University between 1996 and 2002. The main goal of the SKIPPER project was to demonstrate the appli- cability of skeleton-based parallel programming techniques to the fast prototyping of reactive vision applications. This project has produced several versions of a full-fledged integrated pa- rallel programming environment (PPE). These PPEs have been used to implement realistic vi- sion applications, such as road following or vehicle tracking for assisted driving, on embedded parallel platforms embarked on semi-autonomous vehicles. All versions of SKIPPER share a common front-end and reperto…
Fast prototyping of parallel image processing applications using functional skeletons
We present SKiPPER, a software dedicated to the fast prototyping of vision algorithms on MIMD/DM platforms. This software is based upon the concept of algorithmic skeletons, i.e. higher order program constructs encapsulating recurring forms of parallel computations and hiding their low-level implementation details. Examples of such skeletons in low- to mid-level image processing include such as geometric decompositions, data or task farming. Each skeleton is given an architecture-independant functional (but executable) specification, a portable implementation as a process template and an analytic performance model. The source program is a purely functional specification of the algorithm in …
Optimisation conjointe de la taille de stockage et des performances de modèles de classification pour l’authentification de visages
International audience
Dark Count rate measurement in Geiger mode and simulation of a photodiode array, with CMOS 0.35 technology and transistor quenching.
International audience; Some decades ago single photon detection used to be the terrain of photomultiplier tube (PMT), thanks to its characteristics of sensitivity and speed. However, PMT has several disadvantages such as low quantum efficiency, overall dimensions, and cost, making them unsuitable for compact design of integrated systems. So, the past decade has seen a dramatic increase in interest in new integrated single-photon detectors called Single-Photon Avalanche Diodes (SPAD) or Geiger-mode APD. SPAD detectors fabricated in a standard CMOS technology feature both single-photon sensitivity, and excellent timing resolution, while guarantying a high integration. SPAD are working in ava…
Design and Characterization of Automated Color Sensors System
International audience; The paper presents a color sensor system that can process light reflected from a surface and produce a digital output representing the color of the surface. The end-user interface circuit requires only a 3-bit pseudo flash analog-to-digital converter (ADC) in place of the conventional/typical design comprising ADC, digital signal processor and memory. For scalability and compactness, the ADC was designed such that only two comparators were required regardless of the number of color/wavelength to be identified. The complete system design has been implemented in hardware (bread board) and fully characterized. The ADC achieved less than 0.1 LSB for both INL and DNL. The…
A starting point for real-time human action detection
Analyzing videos of human actions involves understanding the spatial and temporal context of the scenes. State-of-the-art approaches have demonstrated impressive results using Convolution Neural Networks (CNNs). However, most of them operate in a non-real-time, offline fashion and are not well-equipped for many emerging real-world scenarios, such as autonomous driving and public surveillance. In addition, they are computationally demanding to be deployed on devices with limited power resources (e.g., embedded systems). This paper reviews state-of-the-art methods based on CNN for human action detection and related topics. Following that, we propose an initial framework to efficiently address…
Overview of ghost correction for HDR video stream generation
International audience; Most digital cameras use low dynamic range image sensors, these LDR sensors can capture only a limited luminance dynamic range of the scene[1], to about two orders of magnitude (about 256 to 1024 levels). However, the dynamic range of real-world scenes varies over several orders of magnitude (10.000 levels). To overcome this limitation, several methods exist for creating high dynamic range (HDR) image (expensive method uses dedicated HDR image sensor and low-cost solutions using a conventional LDR image sensor). Large number of low-cost solutions applies a temporal exposure bracketing. The HDR image may be constructed with a HDR standard method (an additional step ca…
Scene-based noise reduction on a smart camera
International audience; Raw output data from CMOS image sensors tends to exhibit significant noise called Fixed-Pattern Noise (FPN) due to on-die variations between pixel photodetectors. FPN is often corrected by subtracting its value, estimated through calibration, from the sensor's raw signal. This paper introduces an on-line scene-based technique for an improved FPN compensation which does not rely on calibration, and hence is more robust to the dynamic changes in the FPN which may occur slowly over time. Development has been done with a special emphasis on real-time hardware implementation on a FPGA-based smart camera. Experimental results on different scenes are depicted showing that t…
Nouvelle tomographie Compton
International audience
Multi-pixel Geiger mode imager for medical application
International audience; Nowadays, there are two types of sensors to detect the low luminous flux, PMT (Photomultiplier Tube) and Geiger-APD (Geiger Avalanche Photodiode). The domain of Geiger-APD has reached an advanced development in the last years. The basic idea of this structure consists in polarizing an APD in Geiger-mode by applying a voltage beyond its breakdown voltage. In this case of polarization, the APD is working in a special mode and is able to detect the single photon. The theory of detection of single photon using this detector has been invented in the beginning of 90's and developed for detection of low light intensity. By using this kind of photodiode in the Geiger-mode, w…
Learning discrete and continuous regularities in two-dimensional settings
International audience
HDR-ARtiSt: a FPGA-based Smart Camera for High Dynamic Range color video from multiple exposures
International audience; A camera is able to capture only a part of a high dynamic range scene information. The same scene can be fully perceived by the human visual system. This is true especially for real scenes where the difference in light intensity between the dark areas and bright areas is high. The imaging technique which can overcome this problem is called HDR (High Dynamic Range). It produces images from a set of multiple LDR images (Low Dynamic Range), captured with different exposure times. This technique appears as one of the most appropriate and a cheap solution to enhance the dynamic range of captured environments. We developed an FPGA-based smart camera that produces a HDR liv…
Implémentations matérielles d'un système biométrique bimodal
National audience
Low-noise and low power photoreceptor using split-length MSOFET
International audience
Special issue on architectures of smart cameras for real-time applications
Smart cameras are embedded vision systems whose primary function is to produce a semantic understanding of the scene and generate a response in the form of application-specific signals and data. They are autonomous vision systems themselves and can be the building blocks of a more complex smart camera network. They are built around high-performance on-chip and on-board computing and communication infrastructure, combining image sensing, real-time image and video processing, and communications into a single embedded device. They can also be interconnected in networks and cooperate to provide access to many views, enabling more challenging applications in fields like visual control, surveilla…
HDR-ARtiSt: A 1280x1024-pixel Adaptive Real-time Smart camera for High Dynamic Range video
International audience; Standard cameras capture only a fraction of the information that is visible to the human visual system. This is specifically true for natural scenes including areas of low and high illumination due to transitions between sunlit and shaded areas. When capturing such a scene, many cameras are unable to store the full Dynamic Range (DR) resulting in low quality video where details are concealed in shadows or washed out by sunlight.The imaging technique that can overcome this problem is called HDR (High Dynamic Range) imaging. This paper describes a complete smart camera built around a standard off-the-shelf LDR (Low Dynamic Range) sensor and a Virtex 6 FPGA board. This …
A smart camera for High Dynamic Range imaging
International audience; A camera or a video camera is able to capture only a part of a high dynamic range scene information. The same scene can be almost totally perceived by the human visual system. This is true especially for real scenes where the difference in light intensity between the dark areas and bright areas is high. The imaging technique which can overcome this problem is called HDR (High Dynamic Range). It produces images from a set of multiple LDR images (Low Dynamic Range), captured with different exposure times. This technique appears as one of the most appropriate and a cheap solution to enhance the dynamic range of captured environments. We developed an FPGA-based smart cam…
Architecture-Driven Level Set Optimization: From Clustering to Sub-pixel Image Segmentation
Thanks to their effectiveness, active contour models (ACMs) are of great interest for computer vision scientists. The level set methods (LSMs) refer to the class of geometric active contours. Comparing with the other ACMs, in addition to subpixel accuracy, it has the intrinsic ability to automatically handle topological changes. Nevertheless, the LSMs are computationally expensive. A solution for their time consumption problem can be hardware acceleration using some massively parallel devices such as graphics processing units (GPUs). But the question is: which accuracy can we reach while still maintaining an adequate algorithm to massively parallel architecture? In this paper, we attempt to…
Big Data Applications for Disaster Management
International audience; The term "disaster management" comprises both natural and man-made disasters. Highly pervaded with various types of sensors, our environment generates large amounts of data. Thus, big data applications in the field of disaster management should adopt a modular view, going from a component to nation scale. Current research trends mainly aim at integrating component, building, neighborhood and city levels, neglecting the region level for managing disasters. Current research on big data mainly address smart buildings and smart grids, notably in the following areas: energy waste management, prediction and planning of power generation needs (based on smart meter readings,…
Smart camera design for realtime High Dynamic Range imaging
International audience; Many camera sensors suffer from limited dynamic range. The result is that there is a lack of clear details in displayed images and videos. This paper describes our approach to generate high dynamic range (HDR) from an image sequence while modifying exposure times for each new frame. For this purpose, we propose an FPGA-based architecture that can produce a real-time high dynamic range video from successive image acquisition. Our hardware platform is build around a standard low dynamic range CMOS sensor and a Virtex 5 FPGA board. The CMOS sensor is a EV76C560 provided by e2v. This 1.3 Megapixel device offers novel pixel integration/readout modes and embedded image pre…