摘要
灰度不均匀效应广泛存在于现实图像(real-world images)中,这给图像分割带来了很大的挑战,目前许多的图像分割算法都依赖于图像灰度分布均匀这一假设,这严重影响了算法分割现实图像的分割精度。因此文章结合图像的数学模型,提出了一种基于偏差修正的C-V模型,该方法在水平集函数的演化过程中,同时进行图像的分割与偏移场的估计,利用偏移场的估计值来抑制灰度不均匀效应的影响。仿真结果表明,该方法比经典的C-V模型有更高的分割精度,对初始化轮廓曲线以及噪声有较强的鲁棒性。
A real-world image often has intensity inhomogeneity, which severely challenges image segmentation. Currently ,many image segmentation algorithms assume that the intensity distribution is homogeneous, and the assumption badly affects the precision of their real-world image segmentation. Therefore, using the C-V model, we propose what we believe to be a novel algorithm of level set image segmentation, which eliminates the intensity inhomogeneity by correcting the bias field to suppress the intensity inhomogeneity. To verify the performance of our algorithm, we simulate the segmentation of both real-world images and artificial images. The simulation results, given in Figs. 1 through 3 show that our algorithm has better segmentation precision than that of the tradition C-V model, can effective suppress the intensity inhomogeneity and noise interference and segment both the real-world im- age and the artifical image.
出处
《西北工业大学学报》
EI
CAS
CSCD
北大核心
2013年第2期218-222,共5页
Journal of Northwestern Polytechnical University
关键词
图像分割
偏移场修正
灰度不均匀效应
水平集
image segmentation
bias field correction, Chan-Vese model, intensity inhomogeneity, level set
