Video clip retrieval based on incidence matrix and dynamic-step sliding-window

Abstract

Video clip retrieval is one important way for content-based video retrieval. This paper presents a novel approach for video clip retrieval based on incidence matrix and dynamic-step sliding-window. First, we use the incidence matrix to calculate the maximum matching to eliminate dissimilar video clips. And then we design a dynamic-step sliding-window to segment the similar video clips from the video database. Finally, the vision factor, order factor and interference factor are adopted to calculate final similarity of the video clips. Experiment result indicates that our approach effectively speeds up the retrieval process, and also improves the retrieval performance.