This　paper　proposes　a　seam　carving　algorithm　based　on　saliency.　This　algorithm　makes　saliency　detection　to　the　source　image.　Images　are　classified　according　to　the　gray-scale　of　saliency　detection.　Adding　different　energy　protection　methods,　it　can　protect　the　foreground　area　of　a　subject　and　detailed　areas　of　the　edge.　Then　the　cumulative　energy　map　is　calculated.　According　to　the　principle　of　deleting　the　minimum　energy　pixel,　single　pixel-wide　carves　chosen　to　be　deleted　or　copied.　The　experimental　result　shows　that　the　algorithm　avoids　image　restoration　caused　by　extracting　too　many　pixels　in　saliency　area　and　non　equal-ratio　scaling.　It　reduces　the　deformed　structures　caused　by　displacement　of　edge　pixel　and　also　improves　the　integrity　of　algorithm.　©　Springer　International　Publishing　AG　2018.