Analogy-based volume exploration using ellipsoidal Gaussian transfer functions

Abstract

Much effort has been made on multidimensional transfer function, which is designed for effective exploration of 3D scalar datasets. But now, existing solution for designing transfer function typically focuses on exploring volume independently without any prior knowledge. It remains, however, a big challenge for us to reuse the explored knowledge, experience and results in scientific visualization. In this paper, we present a novel technique that employs an analogy-based approach. It aims to facilitate automatic volume exploration for multiple datasets which may share common context or features. The kernel of our approach is using the template scheme. With the introduction of the Gaussian Mixture Model, we adopt this new scheme to modeling, designing and transferring—they are processed in the data histogram space. Then, we integrate this scheme into two-dimensional transfer function design. The result shows that the interesting features can easily be captured with little user workload after adopting our approach.