3D flow features visualization via fuzzy clustering

Abstract

A key approach to visualizing a flow field is to emphasize regions with significant behavior. However, it is difficult to give concrete criteria for classifying feature regions. In this paper, we use a novel framework in which fuzzy sets are used to determine flow features: Fuzzy relationships assess structural properties of features. A fuzzy c-means-like clustering algorithm is used to evaluate the importance of each voxel. Our approach can be readily modified with new fuzzy relationships describing other features of interest to users. We use a multi-resolution approach which displays structural features in greater detail, and represents the background by coarse-grained information. Experiments on synthetic and real datasets show that our framework can highlight significant aspects of the whole flow while avoiding occlusion and clutter. Interactive performance is achieved via a GPU implementation. © 2011 Springer-Verlag.