Parametric model for video content analysis

In this paper, we propose a parametric model for the video content analysis. We use the autoregressive (AR) modeling to model the frame feature sequence over time and make the future analysis in the AR parametric space. Based on our parametric framework, applications of detecting shot boundaries in video sequences, extracting key frames and combined spatial-temporal features for shot classification are proposed. Experiments show that our new parametric framework can present the video content better than the traditional color histogram. (c) 2007 Elsevier B.V. All rights reserved.