Abstract: Aiming at the characteristic of medical volume rendering, this article studied the variety of vector field of data, analyzed the relationship between the grads vector field of the medical volume data set and the sampling frequency, then proposed an adaptive sampling algorithm, in which the volume area that is more to contribute to the final image would be sampled with highfrequency, and in a relative manner, the volume area that is less to contribute to the final image would be sampled with lowfrequency. In this way, the sampling step is adjusted adaptively, and the performance of volume reconstruction is improved. The experiments show that the performance of the volume reconstruction is improved by about 30%~60%, while the quality of image is ensured.