基于区域信息的水平集医学图像分割

摘要

Medical image segmentation is a very important part in medical diagnosis. In order to solve edge leakage, incomplete segmentation and other issues, a level set image segmentation based on area information was proposed. By introducing signed pressure force to replace the traditional stop function,this method not only can effectively slove the initial curve position sensitive issue, but also make the evolution of the curve out of the single direction defects , the direction of the curve is automatically selected through area information. The size of speed can be effectively controlled by the new speed function. When the curve close to the edge of target, the speed is small, so that the curve can avoid the edge of target leakage. The new speed functioncan save time.when the curve far away from the edge of target,the speed is large.%医学图像分割是医学诊断中非常重要的部分，针对医学图像分割往往出现边缘泄漏和不完全分割等问题，本文提出一种基于区域信息的水平集图像分割方法。该方法通过引入符号压力函数来取代传统的停止函数，可以有效地解决曲线初始位置敏感的问题，同时使得演化曲线摆脱了单方向演化的缺陷，演化曲线根据区域信息来选择演化方向。通过构造新的速度函数可以有效地控制速度大小，当演化曲线接近目标边缘时，速度值较小，可以有效避免边缘泄漏。当曲线远离目标边缘时，速度值很大，节省演化所需的时间。