结合全局和局部灰度拟合活动轮廓模型的图像分割算法

摘要

As Local Image Fitting(LIF) model is sensitive to the location of initial curve and can be easily trapped into local minimums,an active contour model based on image region information is proposed.This model uses global and local image information to segment images.Its energy function consists of four terms: local term,area term,length term and penalty term.By incorporating the local image information into the proposed model,the images with intensity inhomogeneity can be efficiently segmented.The introduction of a global indicating function in the area term can speed up the convergence of the proposed model and avoid being trapped into local minima.As the constrained level-set function of the penalty term approaches the signed distance function,the proposed model does not need re-initialization,thus the segmentation time is reduced.In addition,to segment the interested region of an image,narrow-band realization method is given for the proposed model.Experiment results show that,the proposed model is insensitive to the initial contour;its convergent speed is fast;it can accurately segment images with uneven gray intensity,and can segment the interested region of an image.%针对局部图像拟合（Local image fitting,LIF）模型对初始轮廓敏感和容易陷入局部极小的缺点,本文提出了一种基于图像区域信息的活动轮廓模型。本模型同时利用图像全局和局部信息来分割图像,其能量泛函由局部项、面积项、长度项和惩罚项4项组成。局部项将图像局部信息融入到模型中,使该模型能够有效分割灰度不均图像。面积项通过引入一个全局指示函数,加快了模型的收敛速度,且能避免陷入局部极小。惩罚项约束水平集函数逼近符号距离函数,使模型无需重新初始化,减少了分割时间。此外,为了实现对图像中感兴趣区域的分割,本文给出了模型的窄带实现方法。实验结果表明：本文模型对初始轮廓的敏感性低,收敛速度快,能准确分割灰度不均的图像,且其窄带实现方法能够实现对图像中感兴趣区域的分割。