一种改进的连续最大流图像分割模型
An Improved Continuous Max-Flow Image Segmentation Model
李睿 1刘朝霞 2曹姗姗2
作者信息
- 1. 信阳学院数学与统计学院,河南 信阳 464000
- 2. 中央民族大学理学院,北京 100081
- 折叠
摘要
在图像分割中,基于连续最大流模型的快速算法有明显的优势,但分割结果易受参数和步长的影响,过分割会产生大量阶梯效应的伪影,而且纹理特征不明显.文章提出一种先对图像进行预处理的新型最大流分割模型,并给出一种新的参数选取方式.实验结果表明,文章提出的新算法在速度和分割效果上更有优势.
Abstract
In the image segmentation model,fast algorithms based on the continuous maxi-mum flow model have obvious advantages,However,the segmentation results are easily affected by parameters and step-size,and over-segmentation will produce a lot of step artifacts,and the texture features are not obvious.We propose a new maximum flow segmentation model that preprocesses the image first,and gives a new parameter selection method.Experimental results show that our new algorithm has more advantages in speed and segmentation effect.
关键词
连续最大流/最大类间差/增广拉格朗/日算法/交替方向乘子法Key words
continuous max-flow/OTSU/augmented lagrangian algorithm/ADMM引用本文复制引用
出版年
2024
NETL
NSTL
万方数据