摘要
针对用于图像分割的传统单一和组合活动轮廓模型对初始轮廓敏感或者不能处理灰度不均匀图像的问题,提出了一种综合图像的全局、局部等区域信息和边缘信息的新的组合活动轮廓模型。该模型首先将全局灰度拟合能量、局部灰度拟合能量进行线性组合,组合后的能量再与由边缘探测算子构造的长度能量组成新的能量项;然后根据能量主体变化规律调整组合权重,继而用其分成的两种组合活动轮廓模型先后对图像分割;最后利用高斯滤波函数对水平集函数正则化。同时提出了模型转换、停止准则,实现了模型的自动转换和自动停止。对人工合成图像和真实图像的数值实验表明,该模型对噪声、各种初始轮廓均具有较好的鲁棒性,并具备分割灰度不均匀图像的能力。
Seeing that a conventional single or combined active contour model for image segmentation is sensitive to initial contours and can not handle the images with inhomogeneous intensity information, the paper proposes a new relay active contour model (RACM) that combines edge information and region information. Firstly, the new model line- arly combines the local intensity fitting energy and the global intensity fitting energy, and then a new energy term is constructed by using the combined energy and an edge term which is made up of edge detection function. Secondly, the combined weight is adjusted according to the state of the evolution model, which can realize image segmentation with two combined active contours models one by one. At last, the Gaussion kernel function is used to regularize the level set function. The criteria for model conversion and stopping are proposed to implement the automatic con- version and automatical stop of a model. The results of the experiment on synthetic images and real images show the proposed model is robust to initial contours and noises, and can segment the images with inhomogeneous intensity information.
出处
《高技术通讯》
CAS
CSCD
北大核心
2013年第4期421-429,共9页
Chinese High Technology Letters
基金
863计划(2009AA044001)
机器人技术与系统国家重点实验室开放研究(SKLRS-2010-MS-01)资助项目
关键词
活动轮廓模型
边缘探测函数
全局灰度
局部灰度
灰度不均匀
图像分割
active contours model, edge detection function, global intensity information, local intensity information, intensity inhomogeneity, image segmentation