Object　segmentation　is　one　of　key　issues　in　object　based　video　analysis　and　retrieval.　In　this　paper,　a　basketball　segmentation　algorithm　is　proposed.　The　algorithms　involve　a　hierarchical　basketball　detection　algorithm　and　a　motion　tracking　algorithm　based　on　adaptive　object　model.　Basketball　detection　algorithm　is　started　firstly,　only　when　a　basketball　is　detected,　the　tracking　algorithm　is　started　in　the　next　frame.　Moreover,　there　is　a　tracking　result　evaluation　module　in　motion　tracking　algorithm,　if　it　is　true,　the　tracking　algorithm　is　used　in　the　next　frame,　otherwise,　the　detection　module　is　started　again.　The　experimental　results　from　real　basketball　video　confirm　the　efficiency　of　the　proposed　algorithms.