Semi-automatic feature point extraction using one seed point

This study proposes a new semi-automatic feature detection algorithm using one seed point to provide precise searching for feature points. The proposed method is essentially composed of two steps: building search information and searching process. A search graph, containing nodes and its access relationship, provides the candidate points for the search process. A bi-directional, multi-segment search strategy is then proposed to determine the optimized feature path. The cost function is essentially composed of four terms, in which the first two terms are employed to track the nodes of similar maximum curvatures and directions of minimum curvature variation, while the last two terms are employed to stabilize the path. Each of the costs is explained in detail, and examples are presented to show the effect of each of them. In addition, several examples are presented to demonstrate the feasibility of the proposed approach.