1 October 2011 Semantic shape similarity-based contour tracking evaluation
Author Affiliations +
Optical Engineering, 50(10), 107003 (2011). doi:10.1117/1.3633334
One major problem of contour-based tracking is how to evaluate the accuracy of tracking results due to nonrigid and deformative properties of contours. We propose a shape context-based evaluation measure that considers the semantic shape similarity between the tracked contour and ground-truth contour. In addition, a pyramid match kernel is introduced for shape histogram matching, which can effectively deal with the contours with different scales. Experimental results demonstrate, compared to two start-of-art evaluation measures, our measure effectively captures the local shape information and thus is more consistent with human vision.
Xiaoqin Zhang, Wenhan Luo, Wei Li, Weiming Hu, Li Zhao, "Semantic shape similarity-based contour tracking evaluation," Optical Engineering 50(10), 107003 (1 October 2011). https://doi.org/10.1117/1.3633334

Back to Top