《表2 四种分割方法对图4五张图像的分割结果度量》

  提示：宽带有限、当前游客访问压缩模式

本系列图表出处文件名：随高清版一同展现

《GPU和格子玻尔兹曼方法联合加速的水平集模型及其在图像分割中的应用（英文）》

1. 获取 高清版本忘记账户？点击这里登录

1. 下载图表忘记账户？点击这里登录

Fig.8 shows another set of our validation tests，its purpose is similar to Fig.7.From this set of experiments，we find that the LBF model outputs correct（as shown in the third column of the second row of Fig.8）segmentation result only when the initial contour is completely located inside the target.Under the other two initializations（as shown in the first to second columns of the second row of Fig.8），its outputs are strongly interfered by background clutter and wrong segmentation results are obtained.In sharp contrast to this，our model yields the same correct segmentation results（as shown in the sixth row of Fig.8）under all initialization styles.This further proves that our model has strong insensitivity to contour initialization.At the same time，we also give three intermediate results of our model，which are placed in the third to fifth rows of Fig.8.In addition to the evolution results shown in Fig.8，we also give a series of statistical results（as shown in Fig.9 and each column in its layout corresponds to the same column of Fig.8）related to the evolution process，which include:the global mean image at convergence state，and the internal and external means of the regions separated by the target contour are marked at the upper left corner of the image plane；the numerical distribution images of e1and e2（defined by Eq.(7）)at convergence state；the energy variation curves with respect to the evolution time；the evolution process of three-dimensional stacking form.Through this set of statistics，we can better feel this group of verification experiments.