Abstract: Facing the problem of a large number of small blood vessels becoming blurred or missing in sharpened angiography image based on conventional enhancement algorithm, and even generating a lot of background noise, a phase stretching kernel function was developed, which resulted in an angiography image enhancement algorithm, called extended phase stretch transform enhancement algorithm. This algorithm expended the ‘S’ type group delay’s phase filter to the linear group delay phase filter, and it was proved theoretically that the inverse transformation’s phase of this phase stretch was approximately the normalized second-order gradient of the original image. The traditional gradient extremum expression of high frequency features was converted into angle expression, which was more conducive to highlight and enhance the high frequency features in images. In addition, combined with the relative total variation theory, the windowed total variation measure and the windowed inherent variation measure of pixels were used in enhance process, which made the algorithm better capable of highlighting the edge contour and texture, suppressing the fine and disorderly texture and background noise, overcoming the shortcomings of current methods. Matlab software platform was used to conduct experiments on some image data in data sets such as DeepLesion and OASIS, compared with the enhancement algorithm based on traditional phase stretch transform and image sharpening algorithm upon phase consistency, the experimental results showed that the small blood vessels in the enhanced image were much clear and the background noise was effectively suppressed. The average gradient and information entropy increased by about 50%, which demonstrated the superiority and practicability of the algorithm.