Research

Topics Cooperative fusion for tracking of pedestrians from a moving vehicle Radar-Video sensor fusion LIDAR-Radar-Video sensor fusion Radar pedestrian detection using deep learning Multimodal vision 2D multi-modal video fusion for wide-angle environment perception Visible-thermal video enhancement for detection of road users Automotive High Dynamic Range (HDR) imaging Classic multi-exposure HDR reconstruction Intelligent HDR tone mapping for traffic applications Learning-based HDR video reconstruction and tone mapping Efficient multi-sensor data annotation tool Point-cloud processing Fast Low-level Point-cloud processing Point-cloud based Object detection and tracking Environment mapping and odometry Liborg - Lidar based mapping LIDAR based odometry Monocular visual odometry Automotive occupancy mapping Obstacle detection based on 3D imaging Real-time sensor data processing for autonomous vehicles using Quasar - demo video Cooperative fusion for tracking of pedestrians from a moving vehicle Autonomous vehicles need to be able to detect other road users and roadside hazards at all times and in all conditions.

Topics Networked sensors Sensor networks and methods for wellness monitoring of the elderly Collaborative Tracking in Smart Camera Networks Distributed Camera Networks Multi Camera Networks 3D reconstruction using multiple cameras Real-time video mosaicking Scene and human behavior analysis Foreground background segmentation for dynamic camera viewpoints Foreground/background segmentation Automatic analysis of the worker's behaviour Gesture Recognition Behaviour analysis Immersive communication by means of computer vision (iCocoon) Material Analysis using Image Processing Sensor networks and methods for wellness monitoring of the elderly Addressing the challenges of a rapidly-ageing population has become a priority for many Western countries. Our aim is to relieve the pressure from nursing homes’ limited capacity by pursuing the development of an affordable, round-the-clock monitoring solution that can be used in assisted living facilities. This intelligent solution empowers older people to live (semi-) autonomously for a longer period of time by alerting their caregivers when assistance is required.

Topics Real-time monitoring in Additive Manufacturing Real-time monitoring in Additive Manufacturing In additive manufacturing, items are 3D printed layer-by-layer using materials like plastics, polymers, and metals. Unfortunately, instabilities in the printing process can produce defects like cracks, warping, and pores/voids within the printed item. Our goal at IPI is to develop computer vision systems that identify the creation of these defects in real time, then provide the 3D printer with sufficient information to intervene in a way that corrects, or avoids, the defect. Since these defects can occur over a very short amount of time (< 1 ms), our monitoring systems need to operate at very high speeds while also providing accurate results. Melt pool monitoring: GPU-based real-time detection of pore defects using dynamic features and machine learning Using high speed cameras and photodiodes (sampling rates > 20 kHz), we are exploring AI models that can highlight defect creation.

Today's algorithms (e.g., image/video processing, hyperspectral sensor data, ...) require huge amounts of data. For many algorithms, a good computational performance is indispensable for use in practical applications. These applications are often targeted toward a big diversity of devices, such as desktop PCs, tablets, smartphones, mini PCs. To reach a good computational performance, modern GPUs bring speedups of 10x-200x for highly parallel processing tasks, but one main disadvantage is the difficulty of programming: not only does (properly) programming a GPU require an extensive in-depth knowledge of the details of a GPU, the development efforts are usually high, which causes GPUs not easily to be used for research purposes, e.g., for devising and testing of new algorithms. Then, when CPUs and GPUs of different types and models are combined, the development and debugging complexity level further increases. One of our concerns is that training a developer (in academia, sometimes a.k.a. "Ph.

Topics High Dynamic Range imaging Denoising of time-of-flight depth images and sequences Multicamera image fusion Non-local image reconstruction Multiframe superresolution Demosaicing Error Concealment Wavelet-based denoising of images Non-local means denoising of images Restoration of historical videos Joint removal of blocking artifacts and resolution enhancement High Dynamic Range Imaging Conventional display and image capture technology is limited to a narrow range of luminosities and images associated with such systems have been retroactively called Low Dynamic Range images (LDR). A High dynamic Range (HDR) image on the other hand, refers to an image that encodes a greater range of brightness and luminosity than a reference LDR image. HDR images find a use in near-future, more capable display and camera systems. They allow to more faithfully match the visual impression of a scene to human vision compared to LDR images. Recent developments in so-called HDR televisions have yielded prototypes that can display peak luminosities in the range of 6000 nits.

Topics Hyperspectral Image Restoration Hyperspectral Multi-sensor Data Fusion Spatial Information Modelling Image classification in hyperspectral images Hyperspectral Image Restoration Despite advances in sensor technology, hyperspectral (HS) images are inevitably degraded by noise and blur, which can affect information retrieval and content interpretation. Using denoising and deblurring as a preprocessing tool will improve various post-processing tasks, e.g. classification, target detection, unmixing, etc. We propose a novel restoration algorithm for HS images. Our method first uses PCA to decorrelate the HS images and separate the information content from the noise. The first k PCA channels contain most information of the HS image, and the remaining B ? k PCA channels (where B is the number of spectral bands of HS image) mainly contain noise. If deblurring is performed on these noisy and high-dimensional B ? k PCs, then it will amplify the noise of the data cube and cause high computational cost in processing the data, which is undesirable.

Topics Medical image analysis WaVelocity - Cardiovascular Image Analysis Software Robust Segmentation Methods for Aortic Pulse Wave Velocity Measurement Skeletonization and segmentation for cerebral vessel delineation Generalized profiling with application to arteriovenous malformation segmentation Segmentation of lung airways Skeletonization for best path calculation in 3-D MRI images of blood vessels Fast and memory-efficient 3D segmentation and morphology Focal Cortical Dysplasia (FCD) Detection in MRI Diffusion MRI data analysis and processing MRI segmentation of the developing newborn brain Medical image restoration Parallel MRI+compressive sensing Shearlet regularization for compressive sensing MRI Denoising of medical images Medical image/video quality assessment Medical image and video quality assessment Task-based QA using numerical model observers Task-based QA for interventional X-ray sequences DICOM calibration of medical stereoscopic displays Subjective QA in medical imaging Video latency in laparoscopic surgery Novel imaging techniques 3D Microwave tomography for breast tumor detection Biological imaging Automatic plant phenotyping using image analysis Estimation of objects' features in biological images Analysis of high-throughput screening of C.

Video and image quality assessment Quality assessment (QA) consists of measuring the user's subjective opinion of perceived image/video quality, the user's quality preferences, or the utility of the images for a specific task. The goal of QA is to evaluate and compare imaging systems, and to help drive system design (e.g. sensors, image processing, displays). Image Quality (IQ) is application specific. For example, optimizing Quality can mean: Finding the highest amount of compression that results in the smallest visible artifacts (such as blocking or blurring) for a multimedia video compression codec Minimizing perceptual differences between the reference and the output colors of a color-matching algorithm Assessing the effect of LCD temporal response on cancer detection performance ...and each new technology brings new quality trade-offs. We have applied our research to application domains such as broadcast, surveillance, multimedia, and medical, to evaluate technologies such as color correction, camera automation, video denoising, and video compression.

This project includes interdisciplinary work between image processing specialists, mathematicians and art scholars on virtual restoration and analysis of art work. It is conducted in collaboration with Vrije Universiteit Brussel, Department of Art, Music and Theater Sciences of Ghent University, The Flemish Academic Center for Science and Arts (VLAC) of the Royal Flemish Academy (KVAB) and the Mathematics Department, Duke University, USA. Our case study is the Ghent Altarpiece or the Lam Gods, a polyptych, dated by inscription 1432, painted by Jan and Hubert van Eyck. Topics Virtual restoration of art work Material representation in art work Virtual restoration of art work Our goal is to approximate how the painting looked like before ageing and to aid art historical and palaeographical analysis. In order to achieve this, we developed crack detection and inpainting tools. In particular, we are developing innovative and context-aware patch-based inpainting methods to achieve correct and visually pleasing crack filling result.