The invention relates to a full-automatic coronary artery segmentation method. The method comprises the steps of: segmenting out a heart region; carrying out vessel enhancement processing; automatically detecting seed voxels; and segmenting a coronary artery. Firstly, the heart region comprising the coronary artery is segmented out, a segmented image nearly does not include any noisy point, and both segmentation accuracy and efficiency are superior to a result obtained by methods such as an EM algorithm and the like; by carrying out vessel enhancement processing on the heart region segmented out, a coronary image is obviously enhanced; by carrying out automatic detection of the seed voxels on the image subjected to vessel enhancement processing, a seed voxel set is detected out, the problem that a conventional region segmentation method needs manual intervention is solved, and robustness is improved; and finally, by consistency judgment, the coronary artery is segmented out. Relative to a method for carrying out region segmentation only by using a gray value as the standard, the full-automatic coronary artery segmentation method disclosed by the invention dynamically changes a consistent interval in consistency judgment, and not only avoids manual intervention, but also improves accuracy of segmenting the coronary artery.