3D Mesh Saliency Based on Dictionary Learning with Multi-Level Laplacian-Beltrami Operator

Yu Wang, Xingce Wang, Zhongke Wu, Haichuan Zhao

Abstract: Saliency is an important characteristic in 3D analysis, and saliency detection for 3D meshes has been extensively studied in visual computing. However, insufficient feature description poses a significant challenge for 3D mesh saliency maps, which consist with human visual perception that is independent of the surroundings. We propose a multi-level fusion saliency detection method that leverages multiple eigenfunctions of the Laplacian-Beltrami operator(LBO) through a non-linear suppression operator, effectively integrating information across various levels. This method is based on reconstruction error and sparse matrix within the feature space constructed through dictionary learning (DL) at each level. The effectiveness of this method has been validated on the SchellingData, as demonstrated through both visualization and quantization. The method can be widely applied to down-stream tasks such as mesh simplification, object detection and so on.