In　video　surveillance　system,　it　is　a　key　factor　that　extracting　moving　objects　accurately　to　obtain　better　monitoring　effect.　The　shadows　followed　with　moving　objects　may　change　the　object　shapes　or　distort　the　object.　Therefore,　it　become　difficult　to　extract　moving　objects　accurately.　In　this　paper,　According　to　optical　characteristics　of　shadow,　a　shadow　elimination　method　based　on　HSV　color　space　and　texture　local　cross-entropy　is　proposed.　Experiments　show　that　our　method　is　efficient　in　detecting　and　eliminating　shadows.　©2010　IEEE.