An Integrated Framework for Robust and Fast Automatic Video Segmentation

Cai, Xiaodong, Ali, F H and Stipidis, E (2002) An Integrated Framework for Robust and Fast Automatic Video Segmentation. In: London Communications Symposium, University College London, London.

Full text not available from this repository.

Abstract

In this paper, we extend previous work on video segmentation [1, 2] and propose a novel integrated module-based framework for real-time applications which require automatic, precise, and fast-based features. The framework consists of five main modules: sprite generator, key frame identifier, change detector, video object plane (VOP) extractor and post-processor. A universal approach for sprite-object-based adaptive background update and a two-level change detector present the core technique of the framework. The sprite object background provides an easy way to perform segmentation automatically and the possibility of extracting video objects whenever they appear without considering the change of the background. The analysis of statistical parameter for normalized and high order statistics feature offers accurate segmentation results. The used key frames selection technique offers efficient computation. In addition, the framework targets flexible applications with different motion features by using optional modules.

Item Type: Conference or Workshop Item (Paper)
Schools and Departments: School of Engineering and Informatics > Engineering and Design
Depositing User: Xiaodong Cai
Date Deposited: 06 Feb 2012 20:15
Last Modified: 04 Apr 2012 08:40
URI: http://sro.sussex.ac.uk/id/eprint/25031
📧 Request an update