一种基于断点处边缘方向保持假设的闭合轮廓提取方法  

Closed Contour Extraction Based on An Ending Point's Edge Direction Preserving Assumption

作　　者：关涛[1] 周东翔[1] 刘云辉[2] 

机构地区：[1]国防科学技术大学电子科学与工程学院,长沙410073 [2]香港中文大学机械与自动化工程学系

出　　处：《计算机学报》2014年第6期1335-1341,共7页Chinese Journal of Computers

基　　金：国家自然科学基金(61375032)资助~~

摘　　要：文中提出了一种新颖的闭合轮廓提取方法.分析了当GVF Snake模型处理边缘断裂的图像时,存在无法提取目标原始轮廓信息,尤其是目标边角信息的问题.在GVF外力场演化的能量模型的基础上,基于保持断点处边缘原方向的假设,提出了一种新的具有边角保持特性的能量模型,由此模型得到了边角保持GVF(CP-GVF)外力场.CP-GVF外力场解决了当目标轮廓发生断裂时断点对于GVF外力场的影响问题,能够根据断点处的边缘方向信息,以保持该边缘方向的方式恢复目标轮廓中丢失的边角信息,从而恢复这类目标的原始形状.不同边缘结构的仿真图像和真实图像的实验结果验证了算法的性能.This paper presents a novel closed contour extraction method.When dealing with images with broken edges,the GVF Snake model has the drawback of being unable to extract the contour of the object＇s original shape,especially the corner information.Based on the energy model that derives the GVF field and an ending point＇s edge direction preserving assumption,this paper proposes a new Corner Preserving energy minimization model that can derive a new vector field called Corner Preserving GVF（CP-GVF） field.The CP-GVF field can solve the drawback of GVF field in the way that the lost corner information of the original edge map can be recovered according to the edge direction near the ending point.Therefore,the object＇s original shape is recovered in the CP-GVF field.Various synthetic images with different edge structures and real images have been tested to show the validity of the proposed method.

关 键 词：闭合轮廓提取 GVF SNAKE模型 图像分割 形状恢复 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]