A "bag of words" was presented based method for video representation and matching.First

all local features of all video frames were quantized into a dictionary of visual words.Then each sub-shot of the video was represented by a set of visual words.Finally

a revered index of visual words was created to speed the matching process of video clips.This method not only takes local appearance and spatial information into account

but also compresses the representation of video content.Highly competitive experimental results show that our proposed method is more effective and efficient than former methods for video matching in large video dataset.