- ANR granted project: Cognitive Microscopy (2009-onwards) (see Project MICO ANR-10-TECS-015 + Project SPIRIT JCJC #11 JS02 008 01 )
keywords : image analysis, cognitive vision, digitized microscopy
Within the Medical and Natural Image Understanding team within IPAL lab , our research works aim at bringing together image analysis competencies with knowledge management tools. The cognitive prospective in the field of Information Technologies is gaining momentum: cognitive radio, cognitive vision etc. Yet, defining and delineating the meaning of cognitive is still challenging in the research community. The onset of our projects is usually driven by operational need in the medical practice/biological research community and usually related to new capabilities of acquisition/observation devices. Accordingly, standard computer vision issues are handled in our team such as segmentation and classification issues. In addition, we aim at leveraging reasoning and formal knowledge representation advances to outperform standard way of designing image analysis line processing, in particular in the field of bio-imaging. Based on microscopic challenges, we raise new issues about how to enhance the digitized exploration and handling of biological visual data by proposing a new paradigm: cognitive microscopy. At the end of the day, we believe that this new paradigm can be an asset to the Virtual Physiological Human initiatives worldwide for which a systemic approach involving formal knowledge representation, biological models and workflows and data processing tools is required.
Histopathological Whole Slide Images Exploration project
Neural Stem Cell Tracking Project
- Project Leader for Image Analysis and Data Mining Modules within the ANR funded project RNTL-VISAVI (2007-2010)
keywords : image analysis, supervided classification, food industry
Analyse de l'aspect visuel et tactile de batonnets alimentaires lié au comportement alimentaire des volailles dans l'industrie agro-alimentaire : http://www.visavi-project.org/
Analysis of the visual and tactile aspects of aliments related to the fooding behavior of chickens : http://www.visavi-project.org/
Ph.D. Co-Supervision (CIFRE with ITAVI): Rabie Hachemi (defense on the 15th of June,2010, Paris, France) - Awarded the best poster award at the International Conference on Quality Control by Artifical Vision QCAV'2007.
Local collaborator: Thomas Pons, Research Engineer, responsible for platform development and involved in research follow-up as well.
- Project Leader for Computer Vision Modules within the ANR funded project RIAM-REVES (2007-2009) and France Telecom / Orange (2004-2007)
keywords : video sequence analysis, 3D reconstruction, pose estimation, augmented reality, visio-conference and museography
Estimation de pose 3D d'objets a partir du tracking d'objets/patterns géométriques dans des séquences vidéos et de caméras de qualité moyenne
3D Pose estimation from geometric objects/patterns from video sequence analysis and medium quality cameras. A video of the mixed reality exhibition at La Cité des Sciences et de l'Industrie with a new see-through device technology.
Project awarded the prize for Science and Education at Laval Virtual 2010, the leading European meeting for the virtual reality community.
Local collaborators: Sandy Rihana (postdoctoral fellowship), Stéphane Lemaire (master internship), Franck Shen (Ph.D. student)
Coordinator Partner: Dream In Real
- Project funded by CNES-IGN and PAI Bosphorus of French Ministery of Foreign Affairs (2001-2005) and French Ministery of Research (2005-2007)
- keywords : object recognition, structural model, spatial relation, VHR satellite images
Détection automatique de ponts sur des images satellites haute résolution : L'objectif de ce travail est de détecter automatiquement sur les images satellitales (THR à environ 1 mètre de résolution) les objets cartographiques tels les ponts, les ronds-points ... Nous nous orientons à présent vers l'étude de la génération automatique de modèle à partir d'exemples pour obtenir un algorithme de détection capable de s'adapter à tout type de nouvel objet d'intérêt.
Automatic Bridge Detection System in HR Satellite Images : We design systems for automatically detecting bridges in high-resolution (2.5m) panchromatic IKONOS satellite images using texture information, geometric models, and combinations of both. We have explored several methods for combining both approaches to achieve an improved detection rate without sacrificing to false alarms. Validation showed an increase in detection rate from 57\% (geometry-based) and 33\% (texture-based) to 71\% (best combined system).
Related stuff or projects :
Participation au programme national TECHNOVISION/projet ROBIN (2006-2007)
Ph.D. Co-Supervision: Guray Erus (defense on the 12th of June,2008, Paris, France) - 2009-2010: Postdoctoral fellowship at University of Pennsylvania, Section of Biomedical Image Analysis - Last paper: Automated Segmentation of Brain Lesions by Combining Intensity and Spatial Information, Bilwaj Gaonkar, Guray Erus, Nick Bryan, Christos Davatzikos, ISBI'2010
- Project funded by EADS and French Ministery of Research (1997-2001)
keywords : classification, delaunay triangulation, unorganized point set analysis, morphological operators, sterecocopic robotic vision
Interprétation de nuages de points et Reconstruction 3D de scènes d'extérieur par modèles de facettes: L'objectif de notre travail, réalisé en collaboration avec EADS est de fournir à un robot doté de deux caméras non seulement l'information de profondeur de la scène observée point à point mais aussi une interprétation en termes d'objets 3D pour l'aider à naviguer de façon autonome dans un environnement totalement inconnu, et par conséquent sans modèles a priori.
Point set interpretation and 3D reconstruction of outdoor scenes by models based on facets: Our work is made in collaboration with MAtraerospatiale and deals with 3D reconstruction of outdoor scenes from stereoscopic cloud of 3D points. We detect navigable areas as well as obstacles by structuring clouds by the means of Fuzzy K-Means clustering technics. Then, objects are reconstructed by alpha-shapes, which constitute subgraphes of Delaunay triangulation.
Related stuff or projects :
Introduction
Play with Morphological Operators on Meshes :
Code c with GUI Interface + Application/Applet java for morphological filtering based on Delaunay meshes + Plugin imageJ for morphological filtering based on Delaunay meshes
Applet for Partitionning Unstructured 3D Point Cloud