A novel method to get smooth contour in image segmentation process was presented. First, dynamic programming algorithm was adopted to process the least cumulative cost matrix of object image, and the optimum contour was extracted along the direction that the cumulative costs descend fastest. Then the contour was smoothened and fitted by using adaptive cubic B spline, which adjusts the distribution of control points adaptively according to the contour curvature. Experiment results of many images showed that this method could effectively suppress tiny zigzags on contours and could produce smoother contour curve than other previous methods without losing the fine structures of the contours at the same time.