Article ID Journal Published Year Pages File Type
10335916 Computers & Graphics 2011 15 Pages PDF
Abstract
► We present SpeedSeg, a technique that uses pen speed and curvature information to segment hand-drawn pen strokes into lines and arcs. ► SpeedSeg is able to monitor its own performance and improve the computed segmentation by judiciously merging and splitting segments. ► Studies have demonstrated that SpeedSeg's default parameter values work well for a variety of users, even those with no experience using the system. Furthermore, SpeedSeg achieves high accuracy over a wide range of parameter values. ► SpeedSeg can automatically tune itself to achieve optimal performance for individual users: it uses hill climbing to adjust its parameters to maximize segmentation accuracy on a set of training shapes provided by the user. ► SpeedSeg outperformed the state-of-the-art algorithms. On a data set comprised of 1090 shapes, SpeedSeg achieved a precision of 97.8%, a recall of 96.3%, an f-measure of 0.970, and an all-or-nothing accuracy of 80.0%.
Related Topics
Physical Sciences and Engineering Computer Science Computer Graphics and Computer-Aided Design
Authors
, ,