Abstract: A curve segmetation model is presented,which overcomes the unreasonable division of two-dimensional histogram based on the traditional cross segmentation model.The simplified model is also get by multi-line fitting.In the process of solving the threshhold,the algorithn transfer the two-dimentional threshold to one-dimentional threshold.And in order to get the fast calculation,an iterative method and an optimizing search strategy is used.Compared with the cross segmetation model,the curve segmentation model has the better performance of anti-noise.It maintains the better shape of image edge,and the computation of algorithm is greatly reduced.