摘要
数字图像中粘连目标区域的自动、准确分离是图像分析的难点和热点,直接影响着各类图像自动分析系统的整体性能。本文对凹点分析、椭圆拟合和流域变换三类粘连目标分离方法的主要步骤、优缺点和适用性进行了比较分析。分析结果表明:凹点分析法适用于分离轮廓凹陷明显的少数凸性颗粒构成的简单区域;椭圆拟合法往往适用于有粘连的特定颗粒的自动计数;而流域变换对较多数目颗粒构成区域具有较强的自动分离能力,具有可并行处理、分离彻底等优点。这对有涉及粘连颗粒自动分离的图像自动分析系统的设计具有一定的参考意义。
Automated splitting of overlapping convex particles,especially the complex clumps,remains an important and challenging issue in image analysis,which directly affect the holistic performance.We compared and analyzed the primary steps,advantages and disadvantages,and applicability of three kinds of clump splitting methods,i.e.the concavity analysis,ellipse fitting and watershed transform.The results of anlysis indicate that（1） the splitting algorithm based on concavity analysis is limited to the simple cluster composed of few and convex particles;（2） the ellipse fitting separation method is apt for automatic count of few and convex particles in simple clusters;（3） the watershed transform is a powerful and fast algorithm to split clumps,which has the advantages of being parallelizable,always producing a complete division,and especially efficient and effective for clumps consisted of a large number of components with symmetric or regular shapes.This is beneficial for the design of image analysis or pattern recognition system related to cluster splitting.
出处
《中国体视学与图像分析》
2011年第3期237-242,共6页
Chinese Journal of Stereology and Image Analysis
基金
国家自然科学杰出青年基金(30825011)
973项目(2007CB914603)
关键词
图像分析
粘连
重叠
自动分离
凹点分析
椭圆拟合
极限腐蚀
image analysis
conglutination
overlapping
automatic separation
concavity analysis
ellipse fitting
ultimate erosion
