Adaptive discrete curve evolution for shape recognition

In the recent years, contour-based shape representation is an important issue in the object recognition research area. In this paper, a new shape descriptor A-DCE is proposed based on DCE and DP for contour deformation and recognition. Firstly, the object contour is evolved adaptively by DCE to extract the contour information with important visual parts. Secondly, the costing feature descriptor is computed by Shape Contexts. At last, shape similarity is measured by DP algorithm based on SC costing descriptor. The experimental results conducted on MPEG-7, Kimia and Swedish Leaf shape data sets evaluate the robustness of the proposed on deformed target, and the operational efficiency and retrieval accuracy are both improved.