Image Processing and Analysis
Head: Pierrick Coupé

We design novel methods for surfaces and volumes segmentation and reconstruction from images and videos, and for synthetic image rendering with a goal of computer vision methods validation. Our methods are applied to documents, biological and medical images, and natural images.

Research Groups

The Document Group: The DocCreator project, the ACID project

The Document group focuses its research on document image processing and analysis. We studied three main specific domains: automatic evaluation of document image degradation, navigation into large document image databases, and synthetic generation of document images.

Keywords: document image processing and analysis, document image classification, performance evaluation, old documents.
Projects and collaborations: DIGIDOC, I2S, Arkhenum, Gestform, BNF.

The MorphoBoid Group

The MorphoBoid group studies the automatic classification of images based on image landmarks. Applications are the automatic classification of insects from images, and the analysis of relations between the image landmarks and phenotypes.

Keywords: Landmarks automatic detection, classification, 2D segmentation and registration.
Collaborations: Institut de génétique, environnement et protection des plantes (IGEPP), UMR INRA.


The PICTURA group mainly focuses on the development of efficient and advanced image processing methods and their applications in specific domains such as medical images, 2D images and videos. Our expertise includes image segmentation, restoration, reconstruction, classification and editing.

Keywords: 3D MRI segmentation, automatic diagnosis (Alzheimer, autism, etc.), colorization, inpainting, anonymization and de-identification, processing of large databases, machine learning.
Projects and collaborations: ANONYMA, ACORSENS, SEDUCTION, Technicolor, General Electric, McGill University, Polytechnic University of Valencia, INCIA, RMSB, IMB.

The Video Analysis and Indexing Group

The Video Analysis and Indexing group develops methods for motion estimation, indexing and recognition of spatio-temporal scenes. The core techniques include optical flow estimation, content description and indexing with modeling properties of human visual system or recognition of events and objects of interest.

Application domains: Aid in studies of neurodegenerative diseases (Alzheimer and Parkinson), motion estimation in MRI sequences, indexing of big video data (TRECVID, cultural archives), video surveillance.
Projects and collaborations: MexCulture, TECSAN Tacit, Dem@care, Adasis, Mirane, VisionScope, Widmee, SOGETI.

Page mise à jour le 06/01/2020 à 10:35