A Visual Salience Based Hierarchical Shape Decomposition Algorithm

Shape decomposition is an important foundation for shape understanding and analysis.Based on the understanding of the characteristics of human vision,we proposed a measurement of human vision named visual salience which combines the strength of both skeleton and boundary features.And then a hierarchical shape decomposition method is proposed.First,the skeleton of a shape is produced using a robust distance transform based algorithm.Then,all possible partition lines are retrieved by analyzing the junction points on the skeleton.Next,the visual salience and the radius of the inscribed circle of the junction point are used to discard redundant partition lines.The remaining partition lines are selected to decompose the shape.Experiments show that our method satisfies subjective visual perception on shape decomposition and is robust to large shape noise.Furthermore,our method is flexible and produces controllable decomposition results.