The active contour models (ACM) are especially adapted to the complicated medical image segmentation, because of the powerful ability of incorporating prior knowledge. In this paper, the basic principles and the relationship of two kinds of the ACM were introduced briefly. Then several important improvement schemes were reviewed in detail, including the representations of curves, the gradientbased ACM, the regionbased ACM, and the ACM incorporating prior shape information. After that several main applications in medical image segmentation were discussed. The future research work about the ACM was presented in the end.