摘要
在无初始化水平集图像分割模型中,为解决弱边界、噪声图像的过分割,提出综合区域信息与梯度信息的无初始化图像分割方法.该方法综合利用图像梯度信息和区域的灰度信息构造外部能量函数,驱动目标轮廓曲线稳定地收敛于目标边界,同时,引入距离约束函数作为内部能量,保证水平集函数不偏离符号距离函数,避免水平集进化的初始化过程.实验结果表明,该模型对初始轮廓无要求,对弱边界图像和噪声图像有很好的分割效果.
In the model of image segmentation without initialization,the paper proposes a new image segmentation of integrated regional information and gradient information without initialization to avoid over segmentation of image with dim boundary or noise.The method makes use of gradient information of image and regional gray level information to establish external energy function,which drives the target contour to remain steadily within the boundary.Meanwhile,distance constraint function is used as the internal energy to make sure that the level set function does not deviate from the signed distance function,avoiding initialization of level set evolution.Experiments show that the model has no demand for initial contour.It produces ideal effect of segmentation for images with low boundary and noise.
出处
《苏州市职业大学学报》
2011年第1期29-32,79,共5页
Journal of Suzhou Vocational University
关键词
图像分割
水平集
无初始化
区域信息
梯度
image segmentation
level set
without initialization
regional information
gradient
