简易珠子链算法——由近似边界进行边缘改善的一种算法

A Simplified Bead Chain Algorithm: An Algorithm for Boundary Refinement From an Approximate Boundary

作者提出了由近似边界确定物体边界的简易珠子链算法。该算法不仅比已有的珠子链算法和改进珠子链算法简单,而且还能满足精度要求。文中采用了二种算法:(1)均方误差准则;(2)相对位置误差。后一方法可用在不能使用均方误差准则的地方。两种方法求得的结果均一致。经大量试验和误差分析证明:加上约束条件的算法,比不加约束条件的算法要精确;第一次迭代计算所求得的结果优于其后续的计算。作者还以确定肝细胞显微图象某一线粒体的边界为实例,证明了上述论点是正确的,用简易珠子链算法改善物体的边界是行之有效的。

A simplified bead chain algorithm for boundary refinement from an approximate boundary is presented. It is simpler than other bead chain algorithms, as well as satisfied with precision. Two algorithms for error evaluation were described.(1) mean squared error criterion; (2) releative position error, this method could be used for the place where the mean squared error criterion couldn't be used. The results getting from two methods were almost similar.The test and error analysis prove that. (1)algorithms with constrained conditions were more precise than those without constrained conditions; (2)the first iteration produced the best results.Examples for determining mitochondrion boundary of electron micrograph images of liver cell were given in this paper. They testified that not only proceeding viewpoints were correct, but a/so refining object boundaries using simplified bead chain algorithm were effective.