一种利用水平集的图像分割快速GPU并行算法

FAST GPU PARALLEL ALGORITHM FOR IMAGE SEGMENTATION USING LEVEL SET

研究医学生物图像的快速分割问题。针对传统图像分割算法效率低、分割不准确的缺陷,提出一种利用水平集自动演化获得最优图像分割的方法。首先,定义水平集方程,并针对方程中不同分量进行分析,确定以图像灰度为依据的最优化算法;然后通过对差分方程的离散化,定义最优化算法的求解步骤,并使得该最优化计算方法能并行化处理。该方法可以有效地对医学图像进行分割,尤其适合并行化GPU处理,在确保图像分割质量的前提下,极大地提高了运算效率。

The fast segmentation of medical and biomedical images is studied in the paper.Aiming at the defects of traditional image segmentation algorithm in low efficiency and inaccurate,we propose a method which uses the automatic evolution of level set to obtain the optimal image segmentation.Firstly,we define the level set function,and analyse different components in the function as well as determine the optimisation algorithm which takes the grey scale of image as the basis.After that,we define the solution procedure of the optimisation algorithm by discretising the differential function,and enable this algorithm to perform parallelised processing.This method can effectively segment the medical images,in particular,it is suitable for parallelised GPU processing.It greatly improves the operation efficiency in premise of guaranteeing the quality of image segmentation.