Video segmentation with extraction of overlaid objects via Multi-Tier Spatio-Temporal analysis

Yih Haw Jan*, David W. Lin

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

1 Scopus citations


We present an algorithm for automatic segmentation and tracking of video objects for natural video scenes. The algorithm employs low-level spatio-temporal signal processing and the segmentation is based on analysis of apparent motion, interframe pixel value changes, edges, and textural homogeneity of image regions. It is designed to be able to separate multiple overlaid objects that do complicated or relatively fast motion, to handle object deformation, and to address appearance and disappearance of objects. A distinguishing feature of the algorithm is its three-tier structure. The first and lowest tier extracts a set of low-level features from each video frame. The second and middle tier employs these features to effect a segmentation of each video frame into a moving part (called "foreground") and a stationary part (called "background"). The third and highest tier then identifies and tracks the overlaid objects in the foreground based on motion analysis and morphological operations. Experiments on several different kinds of video show that the algorithm can yield reasonably good identification of object boundaries. We also point out several items of potential future work.

Original languageEnglish
Pages (from-to)205-217
Number of pages13
JournalInternational Journal of Electrical Engineering
Issue number3
StatePublished - 1 Aug 2004


  • Automatic video segmentation
  • MPEG-4
  • Multiple object tracking

Fingerprint Dive into the research topics of 'Video segmentation with extraction of overlaid objects via Multi-Tier Spatio-Temporal analysis'. Together they form a unique fingerprint.

Cite this