基于改进水平集和区域生长的轮廓提取方法

Contour extraction method based on improved level set and self-adaptive region growing

以人体大脑海马切片序列为研究对象,提出了基于目标灰度差异的水平集方法,该方法克服了水平集方法用于提取海马轮廓时边界易停留在背景梯度局部极值处的问题。同时为了避免传统区域生长人工定义阈值的盲目性,将水平集分割后的图像的标准差作为阈值进行自适应区域生长,获得了海马的轮廓。通过理论分析与实验验证,提出的方法能够有效地滤除非目标区域对海马目标的干扰,获得了较好的分割效果。

This paper proposed a new level set based on target gray differences to research the human brain hippocampal slice sequences.It solved the boundary staying at the background’s extreme gradient in traditional methods.In order to avoid the blindness in region growing,it put forward the processed image’s standard deviation as the threshold.Then it received the hippocampus’s contour.By theoretical analysis and experiment verification,this method can remove non-target parts from target efficiently,and a better segmentation result is obtained.