The　discrepancies　among　the　spectral　range　of　the　component　bands　of　MHSRRSI　will　result　in　remarkable　differences　of　the　response　extent　to　detect　edges　of　geographical　features.　Considering　the　unique　property　of　each　color　space,　an　improved　Canny　operator　has　been　proposed　which　is　available　to　detect　and　extract　vector　or　scalar　edges　from　MHSRRSI　in　weighted　color　space　(such　as　RGB,　linear　YIQ　and　YUV,　nonlinear　IHS,　and　uniform　CIELUV).　Suggestions　on　feature　selection　and　color　space　suitability　analysis　for　detecting　edges　of　geographical　features　from　MHSRRSI　are　also　discussed.　©　2010　IEEE.