Precise pupil features detection is an important factor for face recognition. This paper presents a robust and accurate algorithm to precisely estimate pupil features: pupil center and pupil radius. First, a pupil parameters estimation step is taken and a weight ring mask is created. Then, a weighted Hough transform is used to precisely extract pupil features. Because of the estimation step, the influence of eyelid and eyelash is largely eliminated. The experimental results show very good robustness and accuracy.