An efficient geometric shape coding and representation approach to obtain both skeleton and contours

We present a novel approach to the coding and representation of shapes using intersections. Our algorithm combines contour and region based approaches. After exploring hitherto unvisited regions of the shape, we find salient features on the boundary in that region. Our method has a number of attractive features. First, it is efficient to compute, since only small yet salient regions are retrieved and analysed. Second, our algorithm generates a contour and region analysis simultaneously. Third, the algorithm is fully adaptive, so we do not need to set parameters, such as the number of features to be found, in advance. Finally, the algorithm can be implemented using familiar data structures including the max heap, circular linked lists and trees. We demonstrate the stability of the algorithm with a few practical image sets and developed an extended version which deals with enclosed shapes.