Video　key　frame　extraction　is　a　type　of　video　abstraction,　which　is　one　of　the　key　problems　in　video　content　indexing　and　retrieval.　Key　frame　extraction　aims　at　finding　a　small　collection　of　salient　images　extracted　from　a　video　sequence　for　visual　content　summarization.　In　this　paper　we　propose　a　video　key　frame　extraction　method　based　on　spatial-temporal　color　distribution.　First　we　construct　a　temporally　maximum　occurrence　frame　which　considers　the　spatial　and　temporal　distribution　of　the　pixels　throughout　the　video　shot.　Then　a　weighted　distance　is　computed　between　frames　in　the　shot　and　the　constructed　reference　frame.　Key　frames　are　extracted　at　the　peaks　of　the　distance　curve　and　can　achieve　high　compression　ratio　and　high　fidelity.