Video search by multi-modal and clustering analysis
Duy-Dinh Le, Fuminori Yamagishi, Shin’ichi Satoh · 2007
We introduce a video search system using multi-modal and clustering analysis. In this system, users can make queries by keywords or by visual contents such as frames or rectangle regions within a frame, the retrieved results are then organized into clusters and only representative frames of these clusters are shown to users. Experiments on TRECVID datasets showed promising results.