Séminaire de Dr Sébastien POULLOT, chercheur de l'UMI JFLI (Japanese-French Laboratory in Informatics) - Date : Mardi 12 mars 2013, 14h00 - Lieu : salle "seminar", B1, MICA Insitute, Hanoi University of Science and Technology

Intervenant :
Dr Sébastien POULLOT, chercheur du Japanese-French Laboratory in Informatics (JFLI) de Tokyo, Japon

Date : mardi 12 mars, 14h00
Lieu : salle "seminar room", 9ème étage, bâtiment B1, Institut MICA, Hanoi University of Science and Technology
Interprète traducteur : le séminaire sera présenté en anglais

Résumé/Abstract:
I will present a new effective method for foreground object segmentation dedicated to streaming videos. Based on the assumption that in videos the objects of attention (foreground objects) are moving while ”the world” (background) is still, we propose to perform a camera work estimation on couple of successive frames, using matching of local features and RANSAC. Then the frames can be aligned and background substraction performed so as to isolate the areas of activity. Each area is potentially a (set of) foreground object(s). Therefore, after a morphological and probabilistic filtering, the grabcut algorithm is applied on remaining locations in order to smartly segment the objects. We first show that this intuitive but naive approach achieved surprisingly competitive performance compared to the state of the art methods. However, we also observe that the method fails when camera work estimation does not work well, because camera work estimation and background region estimation are inextricable processes. To circumvent this, we propose an iterative method between the matching of loc al features and the camera work estimation. The method is fully unsupervised and does not need any prior knowledge. It can handle professional as home made videos. Our simple is fast, thus it gives us a good hope to achieve real time unsupervised foreground object segmentation of streaming videos with parallel cores or GPUs.