Welcome in the Image and Signal Processing Group
Researchers of the ISPGroup use and develop signal and image processing techniques for applications like:
- Signal Acquisition, Compression & Streaming
- (Multiple) Object Tracking
- Content-based Data Retrieval
- Biomedical Signal & Medical Image Processing
- Watermarking of Multi-Dimensional Signals
- Compressed Sensing and Inverse Problem Solving (e.g., in Optics, Astronomy, Medical Imaging)
- Hyperspectral Imaging: Theory, Models, Algorithms and Bio-Medical Applications
- Tomographic Methods in Bio-Medical Applications (CT, CBCT, ...)
- Sparse Signal Representation, Restoration and Processing
- Quantitative Analysis of Microscopy Images
- Machine Learning and Deep Learning for Image Processing and Computer Vision
- December 18th, 2017: Profs C. De Vleeschouwer and L. Jacques have been awarded the 45th "de Boelpaepe" prize (2015-2016) by the Académie royale des sciences, des lettres et des beaux-arts de Belgique (Royal Academy of Science, Letters and Fine Arts of Belgium) for their scientific research on image processing with wavelet analysis and sparsity methods.
- September 15-17, 2017: Congrats to Simon Carbonnelle, member of the ISPGroup, and his collaborators in the "OneBonsai" team which won ArcelorMittal Belgium’s first hackathon in the two categories "Artificial Intelligence" and "Best Overall".
... (older news)
Next and Last ISP Seminars:
- April 18th, Wednesday, 2018, 15h00: (in Shannon Room), Maxime Istasse, "On the recent progress of object detection using neural networks"
- May 28th, Monday, 2018, 14h00: (in Shannon Room), Julien Moreau, TBD
- March 29th, Thursday, 2018, 16h15: (in Shannon Room), Martin Lefebvre (ECS Group), "Hardware architecture for machine learning and image processing"
- March 13th, Tuesday, 2018, 11h: (invited talk) Dr Manuel S. Stein (Chair for Stochastics, Universität Bayreuth and Mathematics Department, Vrije Universiteit Brussel), "High-Performance Wireless Sensing with Low-Complexity Array Measurements"
... (all seminars)
The ISPGroup is active in the following main research topics:
Sensing, Imaging and Rendering: What is the most efficient way to sense a signal or to image reality? How can the sensing stage already "compute" information in a way that "just the information we need" is captured? This is the general research topic that we address here. This covers fields like, inverse problem solving, compressed sensing and compressive imaging, data restoration techniques, hyperspectral and light field imaging, or virtual view point rendering.
Representation and Communication: How can we efficiently represent an image signal, or general data in as few parameter as possible? Which kind of (greedy) algorithm can deduce such parameters? How can this "sparse" representation be transmitted in a compact bitstream, potentially under latency and bitbudget constraints? How can we adapt the forwarded content to semantic user requirements? How can we build personalized summaries of edited video feeds automatically?
Analysis and Interpretation: Video are segmented into semantically meaningful objects, based on texture and motion analysis. Images descriptors are extracted to characterize visual contents. Targets of interest are detected, recognized, and tracked to understand behaviors in natural scenes. Application domains include autonomous production of visual reports (e.g. for team sport events), but also intelligent vision in surveillance, or cells images analysis in biology.
Medical Image Processing: Our group develops image & signal processing tools and models for the use in various medical contexts, including radiotherapy, proton therapy, brachytherapy, surgery, EEG analysis, kinematic, diffusion tensor imaging, etc.
Focus on a randomly picked research topic:
Online Convolutional Dictionary Learning for Multimodal Imaging: Computational imaging methods that can exploit multiple modalities have the potential to enhance the capabilities of traditional sensing systems. In this work, we propose a new method that reconstructs multimodal images from their linear measurements by exploiting redundancies across different modalities. Our method combines a convolutional group-sparse representation of images with TV regularization for high-quality multimodal imaging. We develop an online algorithm that enables the unsupervised learning of convolutional dictionaries on large-scale datasets that are typical in such applications.