基于交互式图论的目标边缘检测算法

针对传统边缘检测算法无法准确提取目标及其边缘的问题，基于交互式图论的最大流/最小割理论提出了一种新的边缘检测算法，设计了一种新的代价函数OE_COST 目标边缘代价函数；通过建立图割模型，能够在分割出目标的同时提取出目标边缘。算法通过交互式选择背景及目标像素集合作为硬性约束，通过图像特征（如灰度级、空间信息等）建立代价函数作为软性约束，同时施加软硬约束达到提取目标边缘的目的。实验结果表明，本算法可以准确提取出目标及其边缘轮廓。

A novel target edge detection algorithm based on interactive graph theory

As the traditional algorithm for edge detection cannot extract the targets and edges accurately, based on the max flow/min-cut theory of interactive graph theory, a novel edge detection method is proposed and a new cost-function, named Objects Edge Cost (OE_COST), is designed. The proposed algorithm establishes the graph cut mode to cut the objects and get their edges. The algorithm can extract the edges of targets through combining the hard constraints and soft constraints. The hard constraints can be got by selecting background seeds and object seeds interactively, while the soft constraints are based on the properties of images such as gray levels and space information. Experimental results show that the proposed algorithm can obtain the objects and edges exactly.