Liver segmentation is an important step in computer-aided diagnosis, treatment and surgery of liver diseases. A liver segmentation method based on spatial fuzzy C-means and graph cuts is proposed. Firstly, in order to remove the influence of adjacent organs and tissues on liver segmentation, the spine and ribs are removed from original CT images by thresholding, projection method and 3D region growing, and the right kidney is removed by K-means and binary morphological reconstruction method. Then, liver is segmented by spatial fuzzy C-means from the initial liver slice. The remaining slices are segmented iteratively by graph cuts based on the spatial, shape and gray scale characteristics of CT volumes. Finally, the inferior vena cava is removed by morphological operations and anatomical knowledge. The experimental results show that the proposed method can obtain better segmentation performance than those of other similar methods.