The　paper　proposed　a　method　that　model　the　shift　of　visual　attended　region　based　on　region　growing　in　gray　image.　It　applied　image　segmentation　by　leader　pixels　and　resulted　in　different　regions　without　intersection.　Then　saliency　values　associated　to　these　regions　were　computed　and　sorted.　The　shift　was　realized　by　these　sorted　saliency　value.　The　experimental　results　show　that　the　method　can　model　and　predict　the　shift　of　possible　meaningful　visual　attended　region.　It＇s　significant　for　modeling　the　shift　of　objects　attended　by　eyes　and　promoting　computer　vision.