Owing to the complex morphology of the blood vessel boundary, the stenosis detection with pixel level cannot reflect the details effectively. We propose a sub-pixel automatic detection method of the blood vessel stenosis based on the digital subtraction angiography (DSA), which can identify the location and the degree of the stenosis more accurately and can obtain accurate quantitative results of the stenosis through sub-pixel analysis. Firstly, we extract the central axis of the blood vessel based on the adaptive multi-scale filtering and morphological operations. Secondly, we perform sub-pixel level detection of the blood vessel by the rotation invariance of Zernike moments. Finally, we quantize the diameter using the diameter measurement algorithm based on dynamical ball. Thus, the sub-pixel automatic detection of blood vessel stenosis based on DSA is realized.