Shape matching by part alignment using extended chordal axis transform

摘要: One of the main challenges in shape matching is overcoming intra-class variation where objects that are conceptually similar have significant geometric dissimilarity. The key to a solution around this problem incorporating structure object descriptor which can be described by connectivity graph customarily extracted from its skeleton. In slightly different perspective, may also viewed as arrangement protruding parts along boundary. This does not only convey part's ordering anti clockwise direction, but these on levels detail. paper, we propose method estimates distance between two conducting part-to-part analysis their visual parts. We start skeleton-based segmentation inspired Chordal Axis Transform. Then, extract segments represent silhouette varied Each one feature vector. A thus vectors addition angular and linear proximities each other. Using dynamic programming, our algorithm finds minimal cost correspondence Our experimental evaluations validate proposition part allows conceptual precisely dissimilar shapes. HighlightsNew concepts employed skeletonization 2D shapes.Experimentally weighing properties salience measure.Shape retrieval measures rather than boundary points.A new approach alignment based correspondence.