基于改进GAC模型的二值水平集前列腺超声图像自动分割算法

Automatic segmentation of prostate ultrasound images using binary level set based on an approved GAC model

提出了一种基于改进测地线主动轮廓(geodesic active contour,GAC)的自动分割算法.首先通过结合径向浅浮槽和区域填充算法得到滤波后图像的大致轮廓,然后通过构造基于区域信息的符号压力函数代替边界停止函数,并且加入了基于边界梯度信息的能量项,有效地克服了弱边界的问题.该模型用二值水平集方法实现,使算法的稳定性更高,计算量大大降低.对前列腺直肠超声图像的实验结果表明:本算法迭代收敛速度快,有效避免了边界泄露问题.

An automatic segmentation algorithm of prostate ultrasound image was proposed using an approved GAC model. First, a coarse contour was obtained by radial bas-relief and region fill algorithm to the filtered image. The problem of weak edges was efficiently solved by presenting a new region-based signed pressure forces function to replace the edge stopping function and incorporating an energy function based on boundary gradient information. The model was implemented by binary level set function, which reduces the expensive computational cost of re-initialization of the conventional level set. Experimental results with several prostate transrectal ultrasound images show that the proposed algorithm has a fast convergence speed compared with conventional GAC models and can avoid the problem of edge leaking effectively.