The　Moving　Total　Least　Squares　(MTLS)　method　has　been　developed　to　fit　for　the　measurement　data　contaminated　with　errors.　Different　from　the　moving　least　squares　method　which　only　takes　into　account　the　error　of　dependent　variable,　MTLS　method　considers　the　errors　of　all　the　variables,　which　determines　the　local　approximants　in　the　sense　of　the　orthogonal　direction.　MTLS　method　is　more　reasonable　than　MLS　method　for　dealing　with　errors-in-variables　(EIV)　model.　But　due　to　the　construction　way　of　local　approximants,　it　is　time　consuming　and　difficult　to　change　the　order　of　basis　function.　This　paper　presents　an　Improved　Moving　Total　Least　Squares　(IMTLS)　method,　in　which　Total　Least　Square　(TLS)　based　on　singular　value　decomposition　is　introduced　to　the　local　approximants.　IMTLS　achieves　the　parameter　estimation　of　local　approximants　using　weight　matrix　instead　of　weight　function　with　compact　support　in　MTLS　method.　Several　examples　of　curve　and　surface　fitting　are　given　to　demonstrate　the　performance　of　IMTLS　method.　©　2015　Elsevier　Ltd.　All　rights　reserved.