AMOS VERSION 1.0 - Semantic Video Object Segmentation Systems


Download binary release

AMOS is an object segmentation and tracking system for general video sources. It combines low level automatic region segmentation with a method for defining and tracking high-level semantic video objects.

A Java interface allows users to select and outline a semantic object in the starting frame of a video. The system then automatically tracks the selected object in the subsequent frames, and produces accurate object boundaries. Regions in both the foreground and the background are tracked, through an innovative way of combining color, edge, and motion information. Users may refine the tracking process at any time, and set parameters optimal for different types of video.

AMOS has been tested on hundreds of video clips, including talking heads, sports, nature scenes, and MPEG test sequences. Excellent performance with respect to accuracy (in terms of subjective and quantitative measures) has been achieved.

The current distribution, implemented entirely in JAVA, includes a binary release of the AMOS system developed at Columbia for MPEG-4 semantic video object segmentation. The binary release is available to the public; the source code is available upon request to ADVENT sponsors.

Publications related to AMOS include the following:

D. Zhong and S.-F. Chang, An Integrated Approach for Content-Based Video Object Segmentation and Retrieval, IEEE Transactions on Circuits and Systems for Video Technology, Vol. 9, No. 8, pp. 1259-1268, Dec. 1999.

For software request or technical/licensing questions, please sent email to Prof. Shih-Fu Chang or Di Zhong.


For problems or questions regarding this web site contact The Web Master.
Last updated: June 12, 2003.