New efficient video similarity searching method with 2-phase filtering

We propose a method of similarity searching for video objects using 2-phase filtering. The method proposed in this paper provides efficient and effective video similarity searching from a set of sub-sequences of video objects given a clip video. It can also find video objects that match to the clip exactly or approximately. Our approach of video search can discard dissimilar video objects in an efficient way and keeping the possible similar video objects. The proposed method has the following features: 1) significant reduction of index size by using DFT (discrete Fourier transform) coefficient values; and 2) realization of efficient similarity searching with 2-phase filtering. First, a basic concept of our method and process flow is shown. Next, details of each stage of the process are introduced. Moreover, some result of preliminary experiments is described in this paper.