A new contour extraction method was presented and applied in MR-CT images to get the contours. The new contour extraction method got the gray threshold firstly. The attribute of image in the criteria of attribute operation was constructed. The selected attribute histogram was structured. The attribute threshold was gotten by the corresponding gray threshold in the attribute histogram. The images were operated by the attribute open and close operation. The useless information in image was removed. Then Roberts operator of classical gradient operators was used to get the image contour. Some important properties such as increment,idempotency,anti-extension and displacement invariability were proposed and proved. This method was experimented on MR-CT images. The results show that the contours of MR-CT images were obtained satisfactorily. The simulating experimental results show that the method getting attribute threshold automatically can retain the important information of image effectively. The new contour extraction method has the strong anti-noise ability and the shapes of contours are kept good.