Spatio temporal key-point detection and video representation in 3D complex wavelet domain

摘要

Technology advances in information technology has been evolving for a long time. With this, the need for effective retrieval of video data has increased manifold. In this paper, we propose spatio temporal key-point based approach for content based video retrieval (CBVR). The proposed framework characterizes videos based on spatio temporal key-points. Most of the existing approaches extract key-points from static video frames, while little work has been done on spatio-temporal key-points. We present a new spatio temporal key-point detector and descriptor using 3D complex wavelet transform. Detected key-points are converted to spatio temporal features to represent videos. Results of empirical analysis are provided to substantiate the retrieving capability of the proposed framework.