|
计算机科学 2006
A New Approach for Gradual Transition Detection Based on Finite-State Automata
|
Abstract:
Video shot boundary detection is an essential step for video browsering and retrieval.Almost every shot boundary detection algorithm detects cut translation without difficulty,but gradual translation is regarded as a hard problem.Gradual translation detection comprises boundary frame determination and boundary frames combination.The former is to determine that a frames is or isn't a gradual boundary frame,and the latter is to determine that a segment which composes bounadary frames and normal framesis or isn't a gradual transition.The approaches focus always on boundary frame determination,and ignore boundary frames combination before.In this paper,we define the tolerance in grandual transition detection,a new conception about boundary frames combination.And we present a new approach for Gradual Transition Detection,which uses Finite-State Automata to obtain proper adaptability and good perform- ance.In TRECVID2004,the approach obtains the best performance in gradual transition detection.