Thursday, March 24th, 2011

Frédéric Precioso, Associate professor at Université de Cergy-Pontoise, Paris, France

Nowadays, providing the user with a system for searching into very large image databases becomes a critical issue of Content-Based Image and Video Retrieval systems (CBIVR). However, bridging the semantic gap between which (semantic) concept(s) the user is looking for and the (digital) content of data is quite difficult. We define new visual data representation and design a (kernel) similarity function on this representation in order to provide a video object retrieval system based on a kernel-based SVM.