A semi-automatic video object segmentation tool is presented for segmenting both still pictures and image sequences. The approach comprises both automatic segmentation algorithms and manual user interaction. The still image segmentation component is comprised of a conventional spatial segmentation algorithm (Recursive Shortest Spanning Tree (RSST)), a hierarchical segmentation representation method (Binary Partition Tree (BPT)), and user interaction. An initial segmentation partition of homogeneous regions is created using RSST. The BPT technique is then used to merge these regions and hierarchically represent the segmentation in a binary tree. The semantic objects are then manually built by selectively clicking on image regions. A video object-tracking component enables image sequence segmentation, and this subsystem is based on motion estimation, spatial segmentation, object projection, region classification, and user interaction. The motion between the previous frame and the current frame is estimated, and the previous object is then projected onto the current partition. A region classification technique is used to determine which regions in the current partition belong to the projected object. User interaction is allowed for object re-initialisation when the segmentation results become inaccurate. The combination of all these components enables offline video sequence segmentation. The results presented on standard test sequences illustrate the potential use of this system for object-based coding and representation of multimedia.