EURASIP Summer School on Remote Sensing and Microscopy Image Processing

• Home
• Application
• Important Dates
• Program
• Poster Presentation
• Accommodation
• Venue/Travel
• Travel Grant
• Lectures
• Teamworks
• Organizers
• Sponsors
• Contact

• Registration Form

• Sponsors
• EURASIP
• IEEE SPS
• University of Pannonia
• University of Szeged
• INRIA

Teamwork

The project works consists of a review of the State of the Art of a particular topic in remote sensing and microscopy. Students are working on these projects in a group of 5 (a total of 3 groups). Each group should search and review the most relevant papers. It is important to identify the major trends, the main types of methodologies applied, and (depending on the topic) datasets, algorithms that are widely used by the community today. The goal is to understand what is going on in the projects' topic.

Each group should prepare:

1. a written review of max. 8 pages + references in IEEE transactions format (IEEEtrans style in Latex/Overleaf) and
2. a 15 min. presentation

Project 1: Deep learning with shape priors

• Proposed by: Ian Jermyn
• Keywords: deep learning, shape, prior, segmentation, classification, geometry
• Abstract: The aim of the project is to present the different ways in which shape information has been combined with deep learning. This might be done by changing the architecture of a neural network model, or by changing the loss function, or by some other method. These techniques can be applied to many types of imagery, including satellite and aerial imagery, and medical, biomedical, and microscopy imagery. The presentation should include the methods used, their similarities and differences, and the algorithms used to learn parameters and perform inference.

Project 2: Data fusion in Earth observation: joint classification of multisensor, multifrequency, multitemporal, or multiresolution remote sensing data

• Proposed by: Gabriele Moser
• Keywords: Data fusion, multimodal, multisensor, multifrequency, multitemporal, multiresolution, semantic segmentation, classification
• Abstract: Current space missions for Earth observation (EO) offer an unprecedented opportunity to collect images of the Earth surface through passive and active sensors, with spatial resolutions ranging from kilometers to a few decimeters, and repetitively in time. On one hand, this wealth of multimodal data is of primary importance in environmental applications (e.g., climate change monitoring, natural disaster management), especially those involving classification and semantic segmentation tasks to map land cover from input remote sensing imagery. On the other hand, it is necessary to develop effective data fusion methods, which can benefit from the complementary information collected by multiple sensors, at multiple times, with multiple spatial resolutions, and in multiple wavelength ranges. In this project, the team will review the literature about data fusion for the semantic segmentation of multimodal remote sensing imagery. The team will be free to focus on one or more fusion tasks (e.g., multisensor, multiresolution, multifrequency, multitemporal fusion) and will review and discuss the main methodological approaches (stochastic modeling, deep learning, kernel machines, ensemble learning, etc.).

Project 3: Target Decomposition Techniques for Full Polarimetric SAR Data

• Proposed by: Avik Bhattacharya
• Keywords: targets, scattering, power components, eigenvalue-eigenvector, model-based, model-free
• Abstract: Target decomposition techniques are pivotal in analyzing full polarimetric Synthetic Aperture Radar (SAR) data, offering a comprehensive understanding of scattering mechanisms and the physical properties of targets. These techniques decompose the first-order information, such as the scattering matrix, or the second-order information, such as the coherency matrix, into simpler components, providing detailed information about the target's characteristics. Several target decomposition methods have been proposed in the literature, offering improved target identification, classification, and interpretation in diverse remote sensing applications.

Project 4: SMLM super-resolution technique for fluorescence microscopy

• Proposed by: Laure Blanc-Féraud
• Keywords: fluorescence microscopy, diffraction, super-resolution, SMLM (Single Molecule Localisation Microscopy), photoactivable fluorophores, molecule localisation, sparse inverse problem
• Abstract:Fluorescence microscopy has revolutionized the way biologists study intracellular structures and the cellular environment. The resolution of fluorescence microscopes is limited by the diffraction of light, and numerous super-resolution techniques have been developed to overcome this limitation. Single Molecule Localisation Microscopy (SMLM) is one such technique. The aim of the project is to present fluorescence microscopy and its limitations in terms of resolution, to introduce the SMLM technique and to describe the numerical methods that enable molecules to be finally localized. Among these methods, the one considering the inverse problem with a sparsity condition will be more particularly described.