A　novel　visual　attention　mechanism　based　on　approach　of　regions　of　interest　(ROIs)　detection　was　presented,　which　segmented　image　regions　with　the　watershed　algorithm.　Image　features　were　extracted　by　sampling　the　center-surround　differences　according　to　biological　visual　attention　mechanism,　which　were　combined　into　a　saliency　map.　Competition　among　salient　points　in　this　map　gave　rise　to　a　single　focus　of　attention　(FOA)　which　was　selected　as　the　seed　point　of　watershed　segmentation.　ROIs　were　extracted　by　combining　salient　regions　with　watershed　segmented　regions.　Based　on　inhibition　of　return　and　proximity,　the　FOA　was　selected　and　shifted　to　measure　the　importance　or　interest　of　the　detected　regions.　The　experimental　results　show　that　the　proposed　method　complies　with　biological　visual　attention　mechanism,　which　is　also　effective　to　reduce　over-segmentation　in　auto-detecting　ROIs　and　performs　well　for　large　objects.