Title :
Shape decomposition based on erosion model
Author :
Kanehara, Fumikazu ; Satoh, Shin´ichi ; Hamada, Takashi
Abstract :
We present a shape decomposition method based on erosion model to extract convex parts from a planar closed shape. Our approach is based on the curvature analysis of shape contours assuming that divisions of a shape generally occurs at concavities on the contour, as if the sea gained on the land along the coast. We call this model as erosion model. The outline of our algorithm is as follows. As a beginning, the vectorized shape contour is extracted from a silhouette image. Next, an erosional vector which represents a force of erosion at each concavity is computed, and then, a boundary which splits the shape is decided. These processes are recursively performed until no division is available. Finally, some parts are merged if necessary. The advantage of our approach is that the meaningful convex part can be obtained efficiently and the decomposition results are theoretically invariant to translation, rotation and scaling
Keywords :
Algorithm design and analysis; Clustering algorithms; Clustering methods; Humans; Partitioning algorithms; Pattern recognition; Read only memory; Shape; Spirals; Stability;
Conference_Titel :
Physics-Based Modeling in Computer Vision, 1995., Proceedings of the Workshop on
Conference_Location :
Cambridge, MA, USA
Print_ISBN :
0-8186-7021-5
DOI :
10.1109/PBMCV.1995.514677