Traitement et Compréhension d’Images
Head : Denis Kouamé
The research domain of the TCI team is image analysis. That is the processing and interpretation of data coming from various image acquisition systems. Our works cover three domains: computer vision, medical imaging and sign language.
The computer vision field develops algorithms and theoretical basis to automatically perform real world data extraction and analysis from digital images. Particularly, we focus on the 3D static scene reconstruction problem observed from one or several viewpoints. On the one hand, it consists of studying the geometrical and photometric relationship specific to the process of forming images and to the priori knowledge on the observed scene. On the other hand, from this relationship we carry out a dense matching between couples of stereoscopic images.
Medical imaging holds a major place in health and care investigation techniques and needs to take into account the complete image production and interpretation chain. Our investigations concern, firstly, the methodological development in image and signal processing applicable or extensible to the principal modalities and deal with image enhancement and quantitative characterization. Secondly we study in detail the problems and applications related to medical imaging modalities; ultrasonic imaging, Optical Coherent Tomography imaging (OCT), Positron Emission Tomography scan imaging (PET-Scan).
In sign language and more generally in vision and gesture-based communication, our research field concerns the analysis and the interpretation of both the posture and the gesture performed by people and captured by vision systems. Thus it is necessary to detect and track various body parts (hand, torso, face), to reconstruct in 3D the posture of characters, to analyze their mimics and to interpret those elements into a linguistic context in order to represent the meaning of the sentence. To take into account the human body variability and the sign language complexity, it is necessary to represent and to take advantage of prior knowledge at each level of the processing chain and to introduce various constrains. Then, on the one hand, our work focuses on the study of dynamic gesture and facial expression models and the sign language modeling. On the other hand, we are interested in using those models by an interpretation system.
This section doesn't contain any article.