In　this　paper,　a　novel　color　image　retrieval　method　based　on　the　regions　of　interest　(ROI)　is　presented.　The　key　technique　involves　the　extraction　of　ROI,　the　description　of　image　feature　and　the　characteristic　matching　algorithm.　To　effectively　extract　ROI,　the　color　domains　were　combined　with　the　HSV　space,　which　fits　the　human　visual　perception　characteristic,　and　then　the　image　was　transformed　by　DWT　to　tune　the　sensitivity　to　its　edge　and　texture.　The　ROI　was　obtained　by　clustering　the　mean　value　and　standard　deviations　of　different　frequency　components　of　the　images.　An　accurate　and　fast　image　retrieval　process　was　realized　by　weighted　color　correlograms　which　effectively　combined　information　from　color　and　space.　The　experimental　results　have　shown　that　the　retrieval　accuracy　and　efficiency　of　the　method　is　greatly　enhanced,　as　compared　to　conventional　methods.　©2009　IEEE.