Mathematical morphology
The Image Processing and Pattern Recognition group contributes to both theoretical and practical aspects of mathematical morphology. Their key achievements notably include the developpement of efficient algorithms for computing and processing hierarchical tree structures (such as the tree of shapes) and their extension to multivariate images and GPU implementation as well as the combination of mathematical morphology with deep learning by developing ways to automatically learn morphological operations through neural networks and robust hierarchical representations.
High-performance image processing
The Image Processing and Pattern Recognition group focuses on developing efficient algorithms with a strong emphasis on execution speed, memory optimization, and ease of use. Notable contributions of the team in the field of high-performance image processing include the continuous development of Pylene, a modern C++ library for mathematical morphology designed for simplicity, efficiency, and versatility, and offering Python bindings as well as state-of-the-art mathematical morphology algorithms optimized for massively parallel architectures like GPUs.
Document analysis and recognition
The document analysis and recognition domain is the historical area of expertise of the Image Processing and Pattern Recognition group, and remains a very active research domain of the team for numerous applications ranging from administrative and identity documents to historical collections.
Medical imagery
The Image Processing and Pattern Recognition group is conducting research in medical imaging since 2017. The team develops efficient algorithms combining mathematical morphology with deep learning for various detection and segmentation tasks, focusing on the detection of small, thin or low-contrast structures.
Space imagery
The Image Processing and Pattern Recognition group adresses challenges in object detection and large-scale data processing on space imagery. Our work involves interdisciplinary collaborations with space science research institutes.
Point cloud processing
The Image Processing and Pattern Recognition group conducts research in point cloud processing, focusing on the development of deep learning models for semantic segmentation and object detection in 3D scans, as well as leveraging exploration robotics for high-precision underwater 3D scene reconstruction.