Persistence Atlas for Critical Point Variability in Ensembles

摘要: This paper presents a new approach for the visualization and analysis of spatial variability features interest represented by critical points in ensemble data. Our framework, called Persistence Atlas , enables dominant patterns points, along with statistics regarding their occurrence ensemble. The persistence atlas represents geometrical domain each pattern form confidence map appearance points. As by-product, our method also provides 2-dimensional layouts entire ensemble, highlighting main trends at global level. is based on notion Map measure density which leverages robustness to noise topological better emphasize salient features. We show how leverage spectral embedding represent members as low-dimensional Euclidean space, where distances between dissimilarities point statistical tasks, such clustering, can be easily carried out. Further, we mandatory leveraged evaluate cluster regions Most steps this framework trivially parallelized efficiently implement them. Extensive experiments demonstrate relevance approach. accuracy provided quantitatively evaluated compared baseline strategy using an off-the-shelf clustering illustrate importance variety real-life datasets, clear feature are identified analyzed. provide lightweight VTK-based C++ implementation that used reproduction purposes.