CAJ | 학술논문

针对现有活动轮廓模型初始化敏感的缺点，提出一种新的基于区域的活动轮廓模型。该模型采用模糊c 均值聚类（FCM）算法对图像进行预分割，将预分割结果二值化为种子标记矩阵，作为下一步水平集演化的初始轮廓，解决了初始化敏感问题；引用RSF（Region-Scalable Fitting）模型的局部区域项作为能量项，提高了分割灰度分布不均匀图像能力；使用高斯滤波方法正则化水平集函数，避免了重新初始化过程，提高了分割效率。实验结果表明：该模型避免了初始化，具有分割结果精确、分割效率高的特点。
침대현유활동륜곽모형초시화민감적결점，제출일충신적기우구역적활동륜곽모형。해모형채용모호c 균치취류（FCM）산법대도상진행예분할，장예분할결과이치화위충자표기구진，작위하일보수평집연화적초시륜곽，해결료초시화민감문제；인용RSF（Region-Scalable Fitting）모형적국부구역항작위능량항，제고료분할회도분포불균균도상능력；사용고사려파방법정칙화수평집함수，피면료중신초시화과정，제고료분할효솔。실험결과표명：해모형피면료초시화，구유분할결과정학、분할효솔고적특점。
Aiming at the defect of existing active contour model that it is sensitive to initialisation,we propose a new region-based active contour model in this paper.This model uses the fuzzy C-means clustering (FCM)algorithm to pre-segment the image,and binarise the results of pre-segmentation to the seed tagged matrix,which solves the problem of initialisation sensitivity;The local area item of the RSF(region-scalable fitting)model is adopted as the energy item,which improves the ability of segmenting the images with inhomogeneous grey distribution;In numerical calculation,Gaussian filtering is utilised to regularise the level set function,which prevents the process of re-initialisation and improves segmentation efficiency.Experimental results show that the proposed model prevents the re-initialisation,it has the characteristics of accurate in segmentation result and high in segmentation efficiency.