On retrieval of flash animations based on visual features

Abstract

Flash is undergoing an explosive spread as a new prevailing media format on the Web. Unfortunately, few research efforts have been devoted to content-based Flash retrieval (CBFR) in IR community, which goes against the utilization of the enormous Flash resources. In this paper, Flash animation in 3-layer architecture is presented after it is segmented based on its visual features to a series of scenes on time-line of the production process. A promising approach of Flash animation retrieval is proposed based on the visual features of scenes and some meta-parameters. An experimental prototype system of Flash animation retrieval with roughly 100,000 Flash animations in total is built. The primary experiment demonstrates the flexibility and the effectiveness of our approach of CBFR. © 2008 Springer-Verlag Berlin Heidelberg.