基于最小方差Snake模型的医学图像分割

传统的参数活动轮廓模型（Snake模型）难于处理自动分割医学图像的弱边界。我们在分析参数活动轮廓和几何活动轮廓模型的基础上，提出最小方差参数活动轮廓模型，并成功应用于医学图像自动分割。该方法将气球力Snake模型中的恒定气球力修改为包含区域信息的变力，以目标和背景两区域具有最小方差为准则，引导轮廓线演化。实验结果表明，该模型对初始轮廓位置不敏感，能实现自动分割。对于带噪声的医学图像，先进行保边界特性的曲率流滤波，然后应用该模型也能取得满意的分割效果。

Medical Image Segmentation Based on The Minimum Variation Snake Model

It is difficult for traditional parametric active contour(Snake) model to deal with automatic segmentation of weak edge medical image.After analyzing snake and geometric active contour model,a minimum variation snake model was proposed and successfully applied to weak edge medical image segmentation.This proposed model replaces constant force in the balloon snake model by variable force incorporating foreground and background two regions information.It drives curve to evolve with the criterion of the minimum variation of foreground and background two regions.Experiments and results have proved that the proposed model is robust to initial contours placements and can segment weak edge medical image automatically.Besides,the testing for segmentation on the noise medical image filtered by curvature flow filter,which preserves edge features,shows a significant effect.