图像分割的自适应快速水平集演化模型  

Adaptive Level Set Evolution Model for Fast Image Segmentation

作　　者：王艳[1] 

机构地区：[1]重庆师范大学数学学院,重庆401331

出　　处：《计算机仿真》2016年第2期269-272,285,共5页Computer Simulation

基　　金：重庆市教育委员会科学技术研究项目(KJ130604);重庆市自然科学基金计划项目(cstc2015jcyj A00019);重庆师范大学基金项目(12XLB034)

摘　　要：在图像分割优化的过程中,为避免传统水平集方法的重新初始化,针对采用无需重新初始化水平集演化模型存在的对轮廓初始化敏感、收敛速度不够快等缺点、首先定义一个均值滤波的可变权系数,并用来替换原模型中面积项的常值权系数,进而得到一个图像分割的自适应快速水平集演化模型。可变权系数在目标内外具有相反符号,使得曲线在演化过程中可以根据图像信息自适应地决定向内或是向外运动,从而允许任意位置定义初始轮廓。实验结果表明,所提模型对轮廓初始化不敏感,而且能够快速分割图像,如对于二值图像一次迭代便可以得到图像分割优化效果。The model of level set evolution without re-initialization can avoid effectively the process of re-initialization in the traditional level set method. However,it is sensitive to contour initialization and the curve sometimes converges slowly. To these issues,an adaptive level set evolution model was proposed in the paper,wherein the constant coefficient associated with the weighted area was modified as an adaptive variable sign coefficient. The coefficient has opposite signs inside and outside the object of interest,thus the curve can automatically proceed in the correct directions,i. e. shrinkage or expansion. Experiments show that the proposed model is free of the initial condition. Besides,the curve converges fast,specially takes only one iteration for binary images.

关 键 词：图像分割 水平集方法 活动轮廓模型 偏微分方程 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]