A segmental deformable model for image segmentation was presented and affine invariants were introduced into the model’s energy function. The contour of the model is deformed hierachically, which keeps the relationship of control points along the contour. Reformative definitions of internal and external energy can reduce computational complexity. In comparison with the classical deformable model, the experimental results demonstrate that the proposed model, which is more effective and more robust to local minima and noise, can achieve better performance for medical image segmentation.