In order to more precisely segment complex microscopic cell image, a new image segmentation method by combination of coarse segmentation and fine segmentation is proposed. Firstly, the coutourlet transform and morphology are used to segment original image coarsely and get the subimages that include the particles. Then ,the Level Set method is employed to locate edge of the particles precisely. The method provides more accurate data for complex microscopic cell automatic recognition system. Taking example for complex urinary sediment image, the experiment results show that the method can segment urinary sediment images effectively and precisely and increasing the performance of urinary sediment particles recognition.