A generic system for semantic video indexing by visual concept

In this paper, we propose the architecture of a generic video indexing system by semantic concepts. It rests mainly on the extraction of visual features deduced from the vicinity of relevant areas and their translation into semantic information. We demonstrate the usefulness of our technique for detecting points of interest in the context of feature extraction and also the using of spatial information combined with bag of words in concept learning and retrieval in large collections of video dataset.