In　order　to　improve　the　embedding　information　capacity　of　picture,　a　steganographic　method　which　based　on　edge　detection　and　exploiting　modification　direction　is　proposed.　According　to　the　human　visual　system　(HVS),　our　method　embeds　secret　data　into　an　edge　region,　no　change　is　happening　in　smooth　areas.　In　the　process　of　this　method,　firstly,　a　cover　image　is　divided　into　3*3　blocks,　and　then　compute　the　gradient　in　four　directions(0,　45.　90　or　135　degree).　For　each　block,　secret　data　are　embedded　into　six　pixels,　the　other　three　pixels　are　used　as　adjuvant.　The　embedded　secret　data　can　be　extracted　from　stego　image　without　referencing　the　cover　image.　The　number　of　embedded　bits　is　dependent　on　the　relation　of　those　six　pixels　and　the　other　three　pixels　accordmg　to　the　magic　matrix.　This　method　can　use　sharper　edge　region　of　each　block　for　data　hiding,　while　preserving　other　smoother　blocks　unchanged.　Hie　improved　performance　of　the　proposed　method　is　proved　by　experimental　results.　Therefore,　the　idea　is　presented　with　more　embedding　capacity　and　got　a　better　pictare　quality.