In　this　paper,　we　present　a　visual　sensitivity　based　low-bit-rate　image　compression　algorithm.　Using　the　idea　that　different　image　regions　have　different　perceptual　significance　relative　to　quality,　the　input　image　is　divided　into　edges,　textures　and　smooth　regions.　For　the　edges,　the　standard　JPEG　algorithm　with　an　appropriate　quantitative　step　is　applied　so　that　the　details　can　be　preserved.　For　the　textures,　the　JPEG　algorithm　is　applied　on　the　down-scale　version.　For　the　smooth　regions,　a　skipping　scheme　is　employed　in　the　compression　process　so　as　to　save　bits.　Experimental　results　show　the　superior　performance　of　our　method　in　terms　of　both　compression　efficiency　and　visual　quality.