Segmentation　of　coronary　artery　intravascular　ultrasound　(IVUS)　images　is　the　first　step　in　assessing　blood　vessel　morphology　and　detecting　possible　atherosclerotic　lesions.　To　overcome　the　disadvantage　of　manual　acquisition　of　initial　contour　in　traditional　Snake　algorithm　and　construction　of　external　force　only　from　gradient　information,　an　improved　Snake　algorithm　is　proposed　for　automatic　outer　membrane　segmentation　in　intravascular　ultrasound　images.　Firstly,　we　design　a　basic　segmentation　algorithm　to　automatically　extract　the　approximate　contour　of　the　first　frame,　and　this　is　iterated　as　the　initial　contour　of　the　Snake　algorithm.　Next,　the　extended　structure　tensor　is　combined　with　the　Snake　algorithm　to　make　it　part　of　the　external　force　of　the　Snake　algorithm,　and　then　the　curve　is　evolved　to　complete　the　segmentation.　To　verify　the　effectiveness　of　the　proposed　method,　this　paper　not　only　shows　the　experimental　results　qualitatively,　but　also　quantitatively　measures　the　RMSE　and　RDD　indices.　Comparative　experiments　show　that　the　proposed　method　not　only　has　higher　accuracy,　but　also　has　better　timeliness.