Coronary　artery　centerline　play　an　important　role　in　diagnosing　cardiovascular　disease,　coronary　artery　registration,　fusion,　biplane　reconstruction　and　so　on.　It　is　still　very　difficult　to　extract　centerline　from　coronary　angiographic　images　with　noise　and　uneven　distribution　agents.　An　centerline　extraction　algorithm　based　on　tracking　and　boundary　information　of　coronary　artery　was　proposed　in　this　paper,　tracking　was　based　on　Hessian　matrix＇s　eigenvalues　and　eigenvectors　and　the　boundary　information　was　obtained　from　narrow　band　level　set　method.　The　bifurcation　points　and　endpoints　of　centerline　which　obtained　from　centerline　extraction　algorithm　were　used　to　segment　coronary　artery　tree.　Simulation　angiography　images　and　clinical　images　that　contain　noise　and　uneven　contrast　media　were　used　to　evaluate　the　performance　of　proposed　algorithm.　The　experimental　results　demonstrated　that　coronary　artery　centerlines　were　extracted　effectively　and　completely　and　coronary　artery　trees　were　divided　into　several　segments　by　using　proposed　algorithm.　This　algorithm　can　work　normally　even　the　images　contains　noise　and　uneven　contrast　media.