Wavelet-Based Feature Vector for Shot Boundary Detection

Abstract

The objective of this paper is to develop a shot boundary detection system to detect shots from a given video. To meet this objective, it is proposed to use feature vector in the process of identifying cut transition and gradual transition. The proposed method analyzes each video frame with its immediate left and right neighbor frame and identifies shot transitions. For each frame comparison, it considers color, edge, motion, and texture features of left and right frames. The proposed method is evaluated on video sequences from the TRECVID video set. The experimental results of the proposed method are compared with the existing methods. It is observed that the proposed method is suitable to perform robust shot boundary detection.