Mesh Saliency Detection Based on Entropy

Aiming at the problem of mesh saliency detection, this paper proposes the entropy based salient model. Firstly, normal vector of each mesh vertex is generated, as the descriptor of each vertex, and each vertex's local neighborhood is chosen. Secondly, histogram is utilized to calculate the probability valuedistribution of the vertexes' normal vector in the neighborhood. Finally, entropy can be used to depict local change of the region, calculate the entropy value of normal vector of all the vertexeswithin the chosen vertex's local neighborhood to get the finalsaliency map. Extensive experimental results demonstrate thatour method is simple and fast, and outperforms many state-of-the-art methods. Furthermore, our method is practical and canbe well integrated into the mesh smooth and mesh simplification.