Fast　and　accurate　estimation　of　background　model　in　video　sequences　is　a　basic　task　in　many　computer　vision　and　video　analysis　applications.　Underwater　vision　is　a　new　area　and　the　background　of　underwater　has　special　quality　such　as　unstable　light　spot,　water　ripple.　To　this　end,　this　paper　proposes　an　algorithm　based　on　Kalman　Filter,　which　is　applied　to　the　estimation　of　dynamic　underwater　background　with　a　static　monitoring　camera　of　swimming　pool＇s　bottom.　Experimental　on　several　underwater　video　sequences　performing　the　model　can　efficiently　adapt　to　the　environmental　of　underwater.　©2010　IEEE.