This　paper　provides　an　effective　web　content-based　image　retrieval　algorithm　by　using　SIFT　(Scale　Invariant　Feature　Transform)　feature.　Different　from　other　existing　text-based　web　image　search　engines,　this　algorithm　can　be　applied　to　content-based　web　image　search　engine　effectively.　SIFT　descriptors,　which　are　invariant　to　image　scaling　and　transformation　and　rotation,　and　partially　invariant　to　illumination　changes　and　affine,　present　the　local　features　of　an　image.　Therefore,　feature　keypoints　saved　as　XML　files　can　be　extracted　more　accurately　by　using　SIFT　than　by　color,　texture,　shape　and　spatial　relations　feature.　To　decrease　unavailable　features　matching,　a　dynamic　probability　function　replaces　the　original　fixed　value　to　determine　the　similarity　distance　of　ROI　(Region　of　interest)　and　database　from　web　training　images.　The　experimental　results　show　that　this　method　improves　the　stability　and　precision　of　image　retrieval.