中文摘要：

结合CT图像的特点，在传统Canny算法的基础上提出了改进的Canny算法．该算法先用基于GCV准则的阈值函数平滑去噪，在得到非极大值抑制图像后用Otsu法自适应设定高低阈值，并用形态学结构元素细化边缘．实验证明，该方法能有效解决常见边缘检测算法对去除噪声和获取精细边缘之间的矛盾，使伪边缘现象大为减少，从而获得了比较理想的边缘检测效果，为提高医生诊断病情的准确率打下了良好的基础．

英文摘要：

In this paper, to address these issues, we propose an improved Canny algorithm based on traditional Canny algorithm. Firstly, we denoise smoothly by threshold function which is based on GCV criteria. Then we set high and low thresholds adaptively with Otsu after getting non--maximum suppression image. Finally, we refine the edge by using morphological structure element. Experiments show that our algorithm can get relatively ideal margin detection effects and avoid the shortages mentioned above, which will lay a good foundation for improving disease diagnose.