With　the　search　algorithm　technology　to　flourish,　the　current　major　search　engine　to　search　the　contents　of　the　letter　has　basically　met　the　needs　of　the　people.　However,　according　to　the　current　accuracy　in　the　image　search,　all　vendors　have　not　given　a　very　good　solution.　This　paper　summarizes　the　solution　adopted　by　current　search　engine　vendor,　and　introduces　a　new　approach　that　attach　content　description　based　on　XML　doc　to　web　images　in　order　to　achieve　converting　unstructured　information　search　into　structured　information　search.　At　the　same　time,　by　injecting　the　Web2.0　feature-Wisdom　of　Crowds,　so　that　the　engine　has　the　characteristics　of　self-learning:　with　the　increase　of　users,　accuracy,　as　well　as　intelligence　of　image　search　engine　will　also　be　increased.